JP7621897B2 - Design plan generation system and method - Google Patents

Description

The present invention relates to a design plan generation system and method, and is suitable for use, for example, in a route generation device that automatically generates routes for infrastructure networks.

For example, when generating routes for infrastructure networks such as power grid transmission routes and communication networks, it is necessary to pursue economic efficiency while observing various constraints, such as those of the natural and social environments. Conventionally, most of these route generation has been done manually. For this reason, generating such routes requires a lot of effort and cost, and there is a demand for technology to automatically generate routes using computers.

However, the constraints and evaluation items that must be considered when generating such routes are numerous and complex, and in many cases are tacit knowledge (know-how) based on the experience and intuition of each expert. In such cases, it is difficult to define the precise constraints and evaluation items required for route generation by computer.

To address this problem, for example, Patent Document 1 discloses an automatic path generation device that creates training data based on a tentative motion path, which is automatically generated using a motion path method algorithm to determine the motion path between multiple motion points along which a robot operates, and an actual motion path between the motion points, which is created by a skilled worker, and generates a trained model that learns the difference between the tentative motion path and the actual motion path using the created training data. In addition, Non-Patent Document 1 discloses an overview of a power transmission line route selection support system that supports the selection of a power transmission line route.

JP 2020-11328 A

“Development of a Power Transmission Line Route Selection Support System: Seeking Environmentally Friendly and Low-cost Power Transmission Line Routes”, Chubu Electric Power Co., Inc., [Retrieved June 14, 2021], Internet〈URL: https://www.chuden.co.jp/resource/seicho_kaihatsu/kaihatsu/kai_library/news/news_2000/news_77_N07705.pdf〉

According to the conventional technology disclosed in Patent Document 1, routes are generated by learning actual movements made by experts, making it possible to generate routes that apply tacit knowledge based on the experience and intuition of experts. However, this does not take into consideration the fact that variations in evaluation items occur depending on the expert. Furthermore, the evaluation items in route generation are treated as a black box, and it can be difficult to obtain a sense of satisfaction with the generated route.

In addition, in the conventional technology disclosed in Non-Patent Document 1, the items to be evaluated in route generation are quantified through a questionnaire survey, but conducting the questionnaire survey requires a lot of effort, and it is difficult to aggregate and calculate multiple implicit evaluation items.

The present invention has been made in consideration of the above points, and aims to propose a design plan generation system and method that can give users a high level of satisfaction by quantifying the evaluation items for each expert and automatically generating a design plan equivalent to that of an expert that corresponds to the evaluation items.

In order to solve the above problem, in the present invention, a design plan generating system for generating a new design plan based on a past design plan includes a classification unit that classifies the past design plans into one or more classes based on the past design plans and performance data including various information necessary for formulating the past design plans, a characteristic learning unit that learns a class classification according to data characteristics based on the performance data of the past design plans belonging to each of the classes classified by the classification unit, an evaluation item calculation unit that calculates, for each of the classes, information that is considered important out of the various information included in the performance data as an evaluation item for that class based on a learning result of the characteristic learning unit, a design plan learning unit that learns the past design plans belonging to that class for each of the classes, the evaluation items for each of the classes calculated by the evaluation item calculation unit, and The new design plan is generated for each of the classes based on the learning results of the past design plans and various information necessary for formulating the new design plan, and a design plan generation unit is provided that calculates the evaluation items in the new design plan for each of the classes, and a result display unit is provided that displays the new design plan for each of the classes generated by the design plan generation unit and the evaluation items in the new design plan , and based on a user's evaluation of the evaluation items for each of the classes calculated by the evaluation item calculation unit and the learning results of the past design plans for each of the classes by the design plan learning unit, the classification of the past design plans by the classification unit, the learning of class classification according to data characteristics by the characteristic learning unit, the calculation of the evaluation items for each of the classes by the evaluation item calculation unit, and the learning of the past design plans by the design plan learning unit are re-executed .
A design plan generating system for generating a new design plan based on a past design plan includes a classification unit for classifying the past design plans into one or more classes based on the past design plans and performance data including various information required for drafting the past design plans, a characteristic learning unit for learning a class classification according to data characteristics based on the performance data of the past design plans belonging to each of the classes classified by the classification unit, an evaluation item calculation unit for calculating, for each of the classes, information considered important among the various information included in the performance data as an evaluation item for the class based on a learning result of the characteristic learning unit, a design plan learning unit for learning, for each of the classes, the past design plans belonging to the class, and the evaluation items for the classes calculated by the evaluation item calculation unit. a design plan generating unit that generates the new design plan for each of the classes based on the items, a learning result of the past design plan for each of the classes by the design plan learning unit, and various information necessary for formulating the new design plan, and calculates the evaluation items in the new design plan for each of the classes; a result display unit that displays the new design plan for each of the classes generated by the design plan generating unit and the evaluation items in the new design plan; and a filtering unit that identifies bad data that satisfies a preset condition from the past design plans and executes a filtering process to remove the identified bad data, and the classification unit classifies the past design plans from which the bad data has been removed by the filtering process into one or more classes.
Further, in a design plan generating system for generating a new design plan based on a past design plan, the system includes a classification unit that classifies the past design plans into one or more classes based on the past design plans and performance data including various information necessary for drafting the past design plans, a characteristic learning unit that learns a class classification according to data characteristics based on the performance data of the past design plans belonging to each of the classes classified by the classification unit, an evaluation item calculation unit that calculates, for each of the classes, information that is considered important among the various information included in the performance data as an evaluation item for the class based on a learning result of the characteristic learning unit, a design plan learning unit that learns the past design plans belonging to the class for each of the classes, the evaluation items for each of the classes calculated by the evaluation item calculation unit, and The new design plan is generated for each of the classes based on a learning result of the past design plan for each class and various information necessary for formulating the new design plan, and a design plan generation unit is provided that calculates the evaluation items in the new design plan for each class, and a result display unit is provided that displays the new design plan for each class generated by the design plan generation unit and the evaluation items in the new design plan, and based on a user's evaluation result for the new design plan and the evaluation items in the new design plan displayed by the result display unit, the classification of the past design plans by the classification unit, the learning of class classification according to data characteristics by the characteristic learning unit, the calculation of the evaluation items for each class by the evaluation item calculation unit, and the learning of the past design plans by the design plan learning unit are re-executed .
Further, in a design plan generating system for generating a new design plan based on a past design plan, the system includes a classification unit that classifies the past design plans into one or more classes based on the past design plans and performance data including various information required for drafting the past design plans, a characteristic learning unit that learns a class classification according to data characteristics based on the performance data of the past design plans belonging to each of the classes classified by the classification unit, an evaluation item calculation unit that calculates, for each of the classes, information that is considered important among the various information included in the performance data as an evaluation item for the class based on a learning result of the characteristic learning unit, a design plan learning unit that learns, for each of the classes, the past design plans belonging to the class, and a design plan generating unit that generates the new design plan for each of the classes based on the evaluation items for the classes, a learning result of the past design plan for each of the classes by the design plan learning unit, and various information necessary for formulating the new design plan, and calculates the evaluation items in the new design plan for each of the classes; a result display unit that displays the new design plan for each of the classes generated by the design plan generating unit and the evaluation items in the new design plan; and an optimization calculation unit that calculates a design plan and evaluation items that satisfy constraint conditions specified by a user through optimization calculation processing, based on the new design plan generated or calculated by the design plan generating unit and the evaluation items in the new design plan.
Further, in a design plan generating system for generating a new design plan based on a past design plan, the system includes a classification unit that classifies the past design plans into one or more classes based on the past design plans and performance data including various information required for drafting the past design plans, a characteristic learning unit that learns a class classification according to data characteristics based on the performance data of the past design plans belonging to each of the classes classified by the classification unit, an evaluation item calculation unit that calculates, for each of the classes, information that is considered important among the various information included in the performance data as an evaluation item for the class based on a learning result of the characteristic learning unit, a design plan learning unit that learns, for each of the classes, the past design plans belonging to the class, and a classification unit that calculates a classification according to the evaluation item calculation unit. a design plan generating unit that generates the new design plan for each of the classes based on the evaluation items for each class, a learning result of the past design plan for each of the classes by the design plan learning unit, and various information necessary for formulating the new design plan, and calculates the evaluation items in the new design plan for each of the classes; a result display unit that displays the new design plan for each of the classes generated by the design plan generating unit and the evaluation items in the new design plan; and a content generating unit that generates practice problems for a user to generate a design plan based on performance data of the past design plans, evaluates the design plan created by the user for the generated practice problems based on the past design plans, and presents the evaluation results to the user.

Also, in the present invention, there is provided a design plan generating method executed in a design plan generating system that generates a new design plan based on a past design plan, the method including: a first step of classifying the past design plans into one or more classes based on the past design plans and performance data including various information required for formulating the past design plans; a second step of learning a class classification according to data characteristics based on the performance data of the past design plans belonging to each of the classified classes; and a second step of calculating, for each of the classes, information that is considered important among the various information included in the performance data as an evaluation item for the class based on a learning result; a second step of learning a previous design plan; a third step of generating the new design plan for each class based on the calculated evaluation items for the class, the learning results of the past design plans for the class, and various information required for formulating the new design plan, and calculating the evaluation items in the new design plan for each class; and a fourth step of displaying the generated new design plan and the evaluation items in the new design plan. The method is so configured that the processing of the first and second steps is re-executed based on a user's evaluation of the calculated evaluation items for the class and the learning results of the past design plans for the class .
A design plan generating method executed in a design plan generating system that generates a new design plan based on a past design plan includes a first step of classifying the past design plans into one or more classes based on the past design plans and performance data including various information required for drafting the past design plans; a second step of learning a class classification according to data characteristics based on the performance data of the past design plans belonging to each of the classified classes, and calculating, based on a learning result, information that is considered important among the various information included in the performance data as an evaluation item for each class, and learning the past design plans belonging to each class; and a third step of generating the new design plan for each of the classes based on the evaluation items for the classes, the learning results of the past design plans for each of the classes, and various information required for formulating the new design plan, and calculating the evaluation items in the new design plan for each of the classes; and a fourth step of displaying the generated new design plan and the evaluation items in the new design plan. In the first step, bad data that satisfies a preset condition is identified from the past design plans, and a filtering process is executed to remove the identified bad data, and the filtered past design plans are classified into one or more classes.
Further, a design plan generating method executed in a design plan generating system that generates a new design plan based on a past design plan includes a first step of classifying the past design plans into one or more classes based on the past design plans and performance data including various information required for planning the past design plans, a second step of learning a class classification according to data characteristics based on the performance data of the past design plans belonging to each of the classified classes, and calculating, for each of the classes, information that is considered important among the various information included in the performance data as an evaluation item for the class based on a learning result, and a third step of calculating, for each of the classes, information that is considered important among the various information included in the performance data as an evaluation item for the class. a second step of learning the design plan of the past design plan for each of the classes; a third step of generating the new design plan for each of the classes based on the calculated evaluation items for the classes, the learning results of the past design plans for the classes, and various information required for formulating the new design plan, and calculating the evaluation items in the new design plan for each of the classes; and a fourth step of displaying the generated new design plan and the evaluation items in the new design plan, and the processing of the first and second steps is re-executed based on the user's evaluation results for the displayed new design plan and the evaluation items in the new design plan.
Further, a design plan generating method executed in a design plan generating system that generates a new design plan based on a past design plan includes a first step of classifying the past design plans into one or more classes based on the past design plans and performance data including various information required for drafting the past design plans, and a second step of learning a class classification according to data characteristics based on the performance data of the past design plans belonging to each of the classified classes, and calculating, for each of the classes, information that is considered important among the various information included in the performance data as an evaluation item for the class based on a learning result, and learning, for each of the classes, the past design plans belonging to the class. a third step of generating the new design plan for each class based on the calculated evaluation items for the class, the learning results of the past design plans for the class, and various information required for formulating the new design plan, and calculating the evaluation items in the new design plan for each class; and a fourth step of displaying the generated new design plan and the evaluation items in the new design plan. In the third step, a design plan and evaluation items that satisfy constraint conditions specified by a user are calculated by an optimization calculation process based on the generated or calculated new design plan and the evaluation items in the new design plan.
Further, a design plan generating method executed in a design plan generating system that generates a new design plan based on a past design plan includes a first step of classifying the past design plans into one or more classes based on the past design plans and performance data including various information required for planning the past design plans; a second step of learning a class classification according to data characteristics based on the performance data of the past design plans belonging to each of the classified classes, and calculating, for each of the classes, information that is considered important among the various information included in the performance data as an evaluation item for the class based on a learning result, and learning the past design plans belonging to each of the classes; and a third step of generating the new design plan for each class based on the evaluation items for the class, the learning results of the past design plans for each class, and various information necessary for formulating the new design plan, and calculating the evaluation items in the new design plan for each class; a fourth step of displaying the generated new design plan and the evaluation items in the new design plan; and a fifth step of generating practice problems for a user to generate a design plan based on performance data of the past design plans, evaluating the design plan created by the user for the generated practice problems based on the past design plans, and presenting the evaluation results to the user.

The design plan generation system and method of the present invention allow the user to easily recognize what information was used as an evaluation item in the generation of a new design plan for each class, and also allows the user to easily confirm the variation in the design plans due to differences in the evaluation items.

The present invention makes it possible to realize a design plan generation system and method that can give users a high level of satisfaction with the generated route.

1 is a block diagram showing a route generating device according to a first embodiment; FIG. 2 is a block diagram showing a logical configuration relating to a learning phase of the route generation device according to the first embodiment. FIG. 2 is a block diagram showing a logical configuration relating to an operation phase of the route generating device according to the first embodiment. 13 is a flowchart showing the flow of processing executed in a learning phase in the route generating device. 11 is a conceptual diagram for explaining the processing content of a data characteristic classification unit. FIG. FIG. 13 is a conceptual diagram illustrating the assignment of a class ID to input data in the operation phase. 13 is a conceptual diagram illustrating the processing content of an evaluation item calculation unit. FIG. 11 is a conceptual diagram for explaining the process of a route candidate generating unit; FIG. FIG. 11 is a diagram showing an example of the configuration of a result display screen according to the first embodiment. FIG. 11 is a block diagram showing a route generating device according to a second embodiment. FIG. 11 is a block diagram showing a logical configuration relating to a learning phase of a route generation device according to a second embodiment. FIG. 11 is a conceptual diagram illustrating a data filtering process. FIG. 11 is a block diagram showing a route generating device according to a third embodiment. FIG. 13 is a block diagram showing a logical configuration relating to a learning phase and an operation phase of a route generation device according to a third embodiment. FIG. 13 is a diagram showing an example of the configuration of a result display screen according to the third embodiment. FIG. 13 is a block diagram showing a route generating device according to a fourth embodiment. FIG. 13 is a block diagram showing a logical configuration relating to an operation phase of a route generating device according to a fourth embodiment. 13 is a flowchart showing the procedure of an optimization calculation process. FIG. 13 is a block diagram showing the configuration of an education system according to a fifth embodiment. FIG. 2 is a block diagram showing a configuration of a content generating device. 13 is a flowchart showing a processing procedure of a training process. 5A to 5C are diagrams illustrating examples of configurations of various screens displayed on a terminal device.

One embodiment of the present invention will be described in detail below with reference to the drawings.

(1) First embodiment (1-1) Configuration of the route generating device according to this embodiment In Fig. 1, the route generating device according to this embodiment is generally designated by 1. This route generating device 1 is composed of a general-purpose computer device equipped with a calculation device 2, a memory 3, a storage device 4, an input device 5 and a display device 6.

The arithmetic unit 2 is a processor that controls the operation of the entire route generation device 1, and is composed of a CPU (Central Processing Unit) or a GPU (Graphic Processing Unit). The arithmetic unit 2 may be composed of a CPU and a GPU, and the CPU and GPU may be switched appropriately by utilizing the characteristics of each program, such as by having the GPU execute programs that allow parallel calculation and the CPU execute programs that require sequential calculation.

The memory 3 is composed of a semiconductor memory such as a RAM (Random Access Memory), and is used as a working memory for the calculation device 2. A data characteristic classification program 10, a data characteristic learning program 11, an evaluation item calculation program 12, a route learning program 13, a route candidate generation program 14, and a result display program 15, which will be described later, are read out from the storage device 4 when necessary and stored and held in this memory 3. The calculation device 2 executes each program read out to the memory 3, thereby executing various processes for the route generation device 1 as a whole, as will be described later.

The storage device 4 is composed of a large-capacity non-volatile storage device such as a hard disk device or SSD (Solid State Drive), and stores various programs and data that needs to be stored for a long period of time. The performance database 20, learning database 21, plan database 22, route candidate database 23, and route evaluation item database 24, which will be described later, are also stored and maintained in this storage device 4.

The input device 5 is composed of, for example, a mouse, a keyboard, and/or a Universal Serial Bus (USB) connector, and is used by the user to input various information and commands to the route generation device 1. The display device 6 is composed of a liquid crystal display, an organic EL (Electro-Luminescence) display, smart glasses or smart goggles, and is used to display various information and screens. Note that instead of providing the input device 5 and the display device 6 separately, a touch panel that integrates these may be used.

(1-2) Route Generation Function According to the Present Embodiment Next, a description will be given of the route generation function installed in the route generation device 1 of the present embodiment. This route generation function is a function that learns routes of an infrastructure network that have been generated by an expert in the past, generates one or more candidates for a route of an infrastructure network to be newly generated (hereinafter, referred to as a route candidate) based on the learning result, and presents each of the generated route candidates to the user.

A series of processes based on such a route generation function consists of a learning phase in which routes of infrastructure networks previously generated by an expert are learned, and an operation phase in which one or more route candidates for the infrastructure network to be newly generated are generated based on the learning results and presented to the user.

Figure 2 shows the logical configuration of the route generation device 1 in the learning phase, and Figure 3 shows the logical configuration of the route generation device 1 in the operation phase. In the figures, the data characteristic classification unit 30, data characteristic learning unit 31, evaluation item calculation unit 32, route learning unit 33, route candidate generation unit 34 and result display unit 35 are functional units that are realized by executing the data characteristic classification program 10, data characteristic learning program 11, evaluation item calculation program 12, route candidate generation program 14 or result display program 15 that are read into memory 3 by the calculation device 2 described above in Figure 1 of the route generation device 1.

In FIG. 2, the performance database 20 is a database in which each route of an infrastructure network that has been generated by an expert in the past and various information required for planning these routes are stored in advance as performance data D1. The performance data D1 is information required when generating routes of infrastructure networks such as, for example, power system transmission routes, communication network routes, transportation networks, and travel routes.

Specifically, when the route of the infrastructure network is a power transmission route for a power system, the performance data D1 includes information on data items such as topographical maps, coordinates of the route and tower locations, coordinates of the locations of buildings, density of buildings, development plans for buildings, coordinates of the locations of ponds, lakes and rivers, flood hazard maps, ecosystem habitat maps, land prices, land availability, contour lines, boring survey results, forest density, wind conditions, construction costs and construction periods.

If the route of the infrastructure network is a transportation network, in addition to the above data items, the performance data D1 also includes information on data items such as the coordinates of the stations, railways, and roads, and the costs associated with the tunnel construction method used. Furthermore, if the route of the infrastructure network is a travel route, the performance data D1 also includes information on data items such as the coordinates of the airports, the length of the airport runways, weather maps, and the wind conditions and weather several hours ahead. It is also possible to create new data items by combining each of the above data items.

The data characteristic classification unit 30 has a function of executing a clustering process for classifying the performance data D1 stored in the performance database 20 into one or more classes that collect similar performance data, and linking the identifier (class ID) of the class to which the performance data D1 is classified to each performance data D1. At this time, the data items used to determine whether or not the data are similar can be selected from any combination of data items from the various data items described above.

Clustering is a method for classifying data based on the similarity between data, and can use, for example, the k-means method, DBSCAN (Density-Based Spatial Clustering of Applications with Noise) method, UMAP (Uniform Manifold Approximation and Projection) method, or KNN (K-Nearest Neighbor) method. The data characteristic classification unit 30 outputs the class ID for each piece of performance data D1 linked as described above to the data characteristic learning unit 31 and the route learning unit 33.

The data characteristic learning unit 31 has a function of learning class classification according to the data characteristics (such as the value of each data item or a combination of values of each data item) of the performance data D1 for each class, using the class ID for each performance data D1 provided by the data characteristic classification unit 30 as teacher data, and outputting the learning results to the evaluation item calculation unit 32. As a learning method at this time, for example, a machine learning method such as a random forest or a support vector machine, or a deep learning method such as a multilayer perceptron (MLP) or a convolutional neural network (CNN) can be used.

The evaluation item calculation unit 32 has a function of using the learning results provided by the data characteristic learning unit 31 to calculate, for each class, one or more data items (hereinafter, these are called evaluation items) that were considered important when generating a route in that class, and the importance ratio of these evaluation items, using a method such as XAI (eXplainable Artificial Intelligence). The evaluation item calculation unit 32 stores the calculated evaluation items and the importance ratio of these evaluation items in the learning database 21.

Here, XAI is a method that can display the characteristics of the learning results, and includes a method of calculating the explainability during the learning process as a method of adding explainability to the learning results. For example, if the learning results are to be added with explainability, a known method of calculating feature importance such as SHAP (Shapley Additive exPlanations) or Grad-CAM (Gradient-weighted Class Activation Mapping) can be used. Furthermore, if the explainability is to be calculated during the learning process, the data characteristic classification unit 30 and the data characteristic learning unit 31 may be treated as a single functional unit, and classification and features may be calculated simultaneously. As such a calculation method, a method based on a decision tree such as a random forest can be used.

Meanwhile, the route learning unit 33 has a function of acquiring each piece of performance data D1 stored in the performance database 20 via the data characteristic classification unit 30, and dividing each piece of acquired performance data D1 into classes based on the class ID for each piece of performance data provided by the data characteristic classification unit 30, and learning a route for each class using the route of the performance data D1 for each class as teacher data. As a learning method at this time, a machine learning method such as a random forest or a support vector machine, or a deep learning method using MLP or CNN can be used. In addition, a regression algorithm or a classification algorithm can be used as a learning algorithm. The route learning unit 33 stores the learning model for each class obtained by such learning (hereinafter, this is called a route learning model) in the learning database 21.

Figure 4 shows the flow of processing within the route generation device 1 in the learning phase. In the route generation device 1, when the input device 5 (Figure 1) is operated and a command is given from the user to learn a route previously generated by an expert, the processing shown in Figure 4 is started, and first, the data characteristics classification unit 30 performs preprocessing, such as normalization and filling in missing values, on each piece of performance data D1 registered in the performance database 20 (S1).

The data characteristic classification unit 30 also executes a clustering process (S2) to classify each piece of performance data D1 that has been subjected to data preprocessing in step S1 into multiple classes as shown in FIG. 5. Note that FIG. 5 shows an example in which the number of classes is "3", but the number of classes may be a number other than 3. In addition, in cases where the number of classes is not determined in advance as in DBSCAN, the number of classes may be determined according to the algorithm. This type of clustering process makes it possible to link a class ID to each piece of performance data D1. The cluster IDs linked in this way represent the variability in knowledge of experts as data, and the process from step S2 to step S6 described below makes it possible to mathematically express the variability in knowledge of experts.

Next, the data characteristic learning unit 31 performs machine learning to classify the performance data D1 using a classification algorithm, using the class IDs of each class of the performance data D1 classified by the clustering process in step S2 as teacher data (S3). By using the learning model obtained by such machine learning (hereinafter, this is called a classification learning model), it is possible to assign an appropriate class ID to unknown input data inputted into the operation phase (assign the unknown input data to an appropriate class), as shown in FIG. 6. By establishing the classification function as a classification learning model in this way, it becomes easier to calculate the feature importance.

Then, the evaluation item calculation unit 32 utilizes the classification learning model obtained in step S3 and uses the XAI method to calculate each evaluation item when generating a route for each performance data D1 belonging to each class for each class as shown in Figure 7, and the ratio of importance of these evaluation items (S4). For example, when generating a route for an infrastructure network, if there are evaluation items such as route length, construction cost, and construction period, the ratio of importance of these evaluation items is calculated for each class. In this way, the importance of each evaluation item that is considered important can be quantified for each class. Note that Figure 7 shows the ratio of importance of each evaluation item for each class as the length of a bar graph.

Then, the evaluation item calculation unit 32 stores the calculated evaluation items for each class and the ratio of the importance of these evaluation items for each class in the learning database 21 (S6).

Meanwhile, in parallel with the processing of the data characteristic learning unit 31 and the evaluation item calculation unit 32, the route learning unit 33 uses each performance data D1 belonging to each class as teacher data and performs machine learning of a route selected according to the value of each evaluation item included in the performance data D1 for each class (S5). By performing learning for each class in this way, it is possible to generate a route learning model that reflects the characteristics of each class.

Then, the route learning unit 33 registers the route learning model for each class thus generated in the learning database 21 in association with the evaluation items of the corresponding class and the ratio of the importance of these evaluation items (S6).

If the information registered in the learning database 21 deviates from what the user expects, the route generation device 1 can be made to re-learn by giving an instruction to introduce new evaluation items and then re-executing the processes in steps S1 to S6 (S7: NO). For example, this is the case when the user is dissatisfied with the number of clusters being higher than expected, or the evaluation items being too subdivided, etc. In this way, the route generation device 1 can be made to learn in a way that is more in line with the user's sensibilities. In particular, in the case of route generation problems in the infrastructure field, where the final decision-making power belongs to humans, such feedback functions can improve the convenience of the route generation device 1.

When the user issues an instruction to the route generation device 1 via the input device 5 to re-learn (S7: YES), the process in the route generation device 1 returns to step S1, and then steps S1 to S6 are executed in the same manner as described above. Then, when a result that satisfies the user is eventually obtained (S7: NO), this series of processes (learning phase) ends.

When steps S1 to S6 are executed repeatedly, the class classification learning model calculated by the data characteristics classification unit 30, the evaluation items for each class and the ratio of the importance of these evaluation items calculated by the data characteristics learning unit 31 for each repeated process, and the route learning model of the route for each class calculated by the route learning unit may all be stored in the learning database 21.

On the other hand, in FIG. 3, the plan database 22 is a database that stores plan data D2 regarding a route of a newly planned infrastructure network. The plan data D2 includes the coordinates of the start point and end point of the planned route, and information on each data item similar to each data item of the performance data D1 described above with reference to FIG. 2.

The route candidate generation unit 34 has a function of generating routes from the start point to the end point in the plan data D2 for each class, as shown in FIG. 8, based on the plan data D2 stored in the plan database 22, using the evaluation items for each class stored in the learning database 21 and the ratio of the importance of these evaluation items for each class, and the route learning model for each class stored in the learning database 21.

Specifically, the route candidate generation unit 34 inputs the plan data D2 into the route learning model of each class, thereby generating a route for each class according to the plan data D2, and stores the generated routes for each class as route candidates in the route candidate database 23.

The route candidate generation unit 34 also reads out the evaluation items for each route candidate (class) and the importance ratios of these evaluation items from the learning database 21, and stores these evaluation items for each route candidate and the importance ratios of these evaluation items in the route evaluation item database 24 in association with the corresponding route candidate.

The result display unit 35 has a function of generating a result display screen that displays, in a predetermined format, route candidates for each class stored in the route candidate database 23, and the evaluation items for each route candidate stored in the route evaluation item database 24 and their importance ratios. The result display unit 35 displays the generated result display screen on a display device.

An example of the configuration of such a result display screen is shown in FIG. 9. This result display screen 40 is configured with route candidate information display areas 41 each corresponding to a route candidate generated by the route candidate generation unit 34 (i.e., each corresponding to each class). Each route candidate information display area 41 displays a route diagram 42 that conceptually shows the path from the start point to the end point of the corresponding route candidate, and a graph 43 that shows the ratio of the importance of each evaluation item in the class corresponding to the route candidate by the length of a bar graph.

(1-3) Effect of this embodiment As described above, the route generation device 1 of this embodiment learns infrastructure network routes that have been generated by experts in the past, generates one or more new infrastructure network route candidates based on the learning results, and presents each of the generated route candidates to the user together with the importance ratio of each evaluation item that is considered important in the class corresponding to the route candidate.

As a result, with this route generation device 1, the user can easily recognize which data items were considered as important evaluation items for each route candidate and the ratio of importance to which these evaluation items were taken into consideration when generating the route candidate. The user can also easily confirm the variation in routes due to differences in evaluation items.

Therefore, according to this embodiment, a user can select from among multiple route candidates a route candidate to actually apply as a route for a new infrastructure network after taking into consideration the evaluation items, thus realizing a route generation device that can give the user a high degree of satisfaction with the generated route.

(2) Second embodiment Figure 10, in which the same reference numerals are assigned to parts corresponding to those in Figure 1, shows a route generating device 50 according to a second embodiment. This route generating device 50 has the same configuration as the route generating device 1 according to the first embodiment, except that a data filtering program 51 is additionally stored in the memory 3.

Figure 11, in which parts corresponding to those in Figure 2 are given the same reference numerals, shows the logical configuration of the route generation device 50 according to this embodiment in the learning phase. The data filtering unit 52 in the figure is a functional unit that is realized when the calculation device 2 (Figure 10) of the route generation device 50 executes the above-mentioned data filtering program 51 (Figure 10) stored in the memory 3 (Figure 10).

The data filtering unit 52 has a function of identifying bad data that satisfies conditions previously specified by the user from among the performance data D1 stored in the performance database 20, reading performance data D1 other than the identified bad data from the performance database 20, and outputting it to the data characteristic classification unit 30.

Here, bad data refers to performance data D1 that should not be used for learning. For example, when designing routes for an infrastructure network, laws and regulations and standards may change from year to year. Furthermore, depending on the period in which the performance data D1 was acquired, the values of data such as ground conditions and weather may change over time. If such performance data D1 is used to cluster the performance data D1, the characteristics of the performance data D1 assigned to each class will not match the current situation, and there is a high possibility that appropriate results will not be obtained in the operation phase.

In the route generating device 50 of this embodiment, the user specifies the conditions for the bad data, such as a period, using a specified screen not shown, and the data filtering unit 52 selectively outputs to the data characteristic classification unit 30 the performance data D1 stored in the performance database 20, other than the performance data D1 that satisfies the conditions specified by the user.

Figure 12 shows the data filtering process by the data filtering unit 52. As shown in the upper part of Figure 12, the performance database 20 stores performance data D1 corresponding to each route of the infrastructure network generated in the past (● corresponds to each individual performance data D1), and the data filtering unit 52 identifies performance data D1 that satisfies the conditions specified by the user from among the performance data D1 (x corresponds to performance data D1 that satisfies the conditions) as shown in the middle part of Figure 12. In addition, the data filtering unit 52 selectively outputs performance data D1 other than the identified performance data D1 to the data characteristic classification unit 30 as shown in the lower part of Figure 12. Thus, the data characteristic classification unit 30 performs clustering of the performance data D1 using only the performance data D1 provided by the data filtering unit 52 in the same manner as in the first embodiment.

In the route generation device 50 of this embodiment, the contents of each process after the data characteristic classification unit 30 in the learning phase and the contents of all processes in the operation phase are the same as those in the first embodiment, so the explanation will be omitted here.

As described above, in the route generation device 50 of this embodiment, bad data that satisfies the conditions specified by the user is removed from the performance data D1 stored in the performance database 20, and then clustering of the performance data D1 is performed, so that appropriate results can be obtained as the clustering result.

Thus, in addition to the effects obtained by the route generation device 1 of the first embodiment, the route generation device 50 can also provide the special effect of being able to provide the user with more reliable route candidates during the operational phase.

(3) Third embodiment Figure 13, in which the same reference numerals are assigned to parts corresponding to those in Figure 1, shows a route generating device 60 according to a third embodiment. This route generating device 60 is configured similarly to the route generating device 1 according to the first embodiment, except that a result evaluation acquisition program 62 is additionally stored in the memory 3, the screen configuration of the result display screen displayed on the display device 6 is different from that of the result display screen 40 described above with reference to Figure 9, and the learning phase process is re-executed based on the user's evaluation results for each route candidate input using the result display screen.

Figure 14, in which parts corresponding to those in Figures 2 and 3 are given the same reference numerals, shows the logical configuration of the route generation device 60 according to this embodiment with respect to the learning phase and the operation phase. The result display unit 63 and the result evaluation acquisition unit 64 in the figure are functional units that are realized when the calculation device 2 (Figure 13) of the route generation device 60 executes the above-mentioned result display program 61 or result evaluation acquisition program 62 stored in the memory 3 (Figure 13), respectively.

The result display unit 63, like the result display unit 35 (Fig. 3) of the first embodiment, has a function of displaying on the display device 6 (Fig. 13) route candidates for each class stored in the route candidate database 23, and the evaluation items and their importance ratios for each route candidate stored in the route evaluation item database 24.

Figure 15, in which parts corresponding to those in Figure 9 are given the same reference numerals, shows an example of the configuration of a result display screen 70 that the result display unit 63 of this embodiment displays on the display device 6. This result display screen 70, like the result display screen 40 (Figure 9) of the first embodiment, is configured to include a route candidate information display area 41 that corresponds to each of the route candidates generated by the route candidate generation unit 34.

In addition to the route map 42 and graph 43 described above in FIG. 9, each route candidate information display area 41 displays a pull-down menu button 71 and an evaluation item setting table 72. Then, on the result display screen 70, the user can evaluate whether the route candidate shown by the route map 42 displayed in the route candidate information display area 41 is appropriate or inappropriate by using a route evaluation setting pull-down menu 73 that is displayed by clicking the pull-down menu button 71 in the same route candidate information display area 41, and the evaluation items used when calculating the route candidate can be changed using the evaluation item setting table 72.

In practice, if the user determines that the route candidate shown by the route map 42 on the result display screen 70 is appropriate from the user's perspective, the user can click and select the word "appropriate" displayed in the route evaluation setting pull-down menu 73 to set the evaluation that the route candidate is appropriate. Also, if the user determines that the route candidate on the result display screen 70 deviates from what the user expects, the user can click and select the word "inappropriate" displayed in the route evaluation setting pull-down menu 73 to set the evaluation that the route candidate is inappropriate.

The evaluation item setting table 72 also displays ON buttons 72A and OFF buttons 72B in the form of toggle buttons corresponding to all data items of the performance data that can be evaluation items. Initially, in the evaluation item setting table 72, the ON buttons 72A are colored a predetermined color for data items that are set as data items (evaluation items) that have been applied to generate the corresponding route candidates, and the OFF buttons 72B are colored the same color for data items that have not been applied.

Then, for each such data item, if the user judges that the data item should be an evaluation item for the corresponding route candidate, the user clicks the ON button 72A to set that the data item should be an evaluation item for the corresponding route candidate, and if the user judges that the data item should not be an evaluation item, the user clicks the OFF button 72B to set that the data item should not be an evaluation item for the corresponding route candidate. At this time, if the clicked ON button 72A or OFF button 72B was not previously displayed in color, the ON button 72A or OFF button 72B is colored the same as the above-mentioned predetermined color, and the color of the other OFF button 72B or ON button 72A is released.

Then, after the user has evaluated each route candidate using the route evaluation setting pull-down menu 73 corresponding to the route candidate as described above, and set the evaluation items for the route candidates using the evaluation item setting table 72, if the user wishes the route generation device 60 to learn these evaluations and settings, the user clicks the additional learning execution button 74.

As a result, the above-mentioned evaluations and settings are fed back to the data characteristics classification unit 30 in the learning phase, and the learning phase processing reflecting the fed back evaluations and settings is executed, and learning is performed according to the evaluations and settings. For example, if a setting is made that limits the number of route candidates or the number of evaluation items, the data characteristics classification unit 30 will execute clustering with a reduced number of clusters.

Then, the operation phase is re-executed based on the learning results of this re-learning, and the result display screen 70 showing the processing results of the re-executed operation phase is generated again by the result display unit 63 and displayed on the display device 6 (Figure 13).

As described above, the route generation device 60 of this embodiment feeds back the user's evaluation results in the operation phase to the learning phase for re-learning, so that route candidates that match the user's intentions can be generated. Therefore, in addition to the effects obtained by the route generation device 1 (FIG. 1) of the first embodiment, the route generation device 60 can also obtain the effect of significantly improving usability when generating routes where the final decision-making power belongs to humans, particularly in the infrastructure field.

(4) Fourth embodiment In the first to third embodiments, the route candidates generated by the route candidate generation unit 34 (FIGS. 3 and 14) are generated using a learning model (classification learning model and route learning model) that is learned based on performance data D1 of routes generated in the past. For this reason, the route generation devices 1, 50, and 60 of the first to third embodiments cannot generate rational route candidates based on specific indicators or constraint conditions. Therefore, in this embodiment, a route generation device that adjusts route candidates generated in the operation phase to more appropriate route candidates using optimization calculations will be described.

Figure 16, in which parts corresponding to those in Figure 1 are given the same reference numerals, shows a route generation device 80 according to the fourth embodiment. This route generation device 80 is configured in the same way as the route generation device 1 (Figure 1) according to the first embodiment, except that an optimization calculation program 81 is additionally stored in the memory 3, and an optimization parameter database 82 and an adjusted route candidate database 83 are additionally stored in the storage device 4.

Figure 17, in which parts corresponding to those in Figure 3 are given the same reference numerals, shows the logical configuration of the route generation device 80 according to this embodiment in the operation phase. The optimization calculation unit 84 in the figure is a functional unit that is realized when the calculation unit 2 (Figure 16) of the route generation device 80 executes the above-mentioned optimization calculation program 81 (Figure 16) stored in the memory 3 (Figure 16).

Here, the optimization parameter database 82 stores in advance data such as the optimization algorithm used in the optimization calculation process executed by the optimization calculation unit 84 as described below, various parameters used in the optimization algorithm, the format of the objective function, and constraint conditions used in the optimization calculation process.

Then, the optimization calculation unit 84 calculates adjusted route candidates that satisfy the various parameters and constraint conditions described above stored in the optimization parameter database 82 based on the route candidates for each class stored in the route candidate database 23 and the evaluation items and the importance ratios of the evaluation items for each route candidate stored in the route evaluation item database 24, and stores the calculated route candidates in the adjusted route candidate database 83.

Thus, the result display unit 35 generates the result display screen 40 described above in FIG. 9 based on each adjusted route candidate stored in the adjusted route candidate database 83 and the evaluation items and the importance ratios of the evaluation items for that route candidate stored in the route evaluation item database 24, and displays the generated result display screen 40 on the display device 6 (FIG. 16).

Figure 18 shows a more specific process flow of the optimization calculation process executed by the optimization calculation unit 84. The optimization calculation unit 84 first reads out all data required for the optimization calculation in step S13 (described later) from the route candidate database 23 (Figure 17), the route evaluation item database 24 (Figure 17), and the optimization parameter database 82 (Figure 17) (S10).

Then, the optimization calculation unit 84 creates an objective function based on each evaluation item of the currently targeted route candidate read from the route evaluation item database 24 in step S10 and the various parameters and the objective function format read from the optimization parameter database 82 in step S10 (S11). The optimization calculation unit 84 also creates constraint conditions for the targeted route candidate based on the constraint conditions read from the optimization parameter database 82 in step S10 (S12).

Next, the optimization calculation unit 84 executes the optimization algorithm read from the optimization parameter database 82 in step S10 (performs optimization calculations) while taking into account the constraint conditions created in step S12 (S13). As a result, a new route candidate is calculated in which the currently targeted route candidate is adjusted while taking into account the constraint conditions created in step S12.

The optimization calculation unit 84 then stores the new route candidate calculated in this manner as an adjusted route candidate in the adjusted route candidate database 83 (Figure 17) (S14), and then ends this optimization calculation process for the currently targeted route candidate.

のベクトルｘ_０で表すことができると仮定する。なおベクトルｘ_０は、対象としているルート候補が当該ルート候補上の各地点をそれぞれ通る確率を順番に並べたベクトルである。 Here, the process of the optimization calculation unit in this embodiment will be specifically described using a mathematical example. Here, the route candidates generated by the route candidate generation unit 34 and stored in the route candidate database 23 are calculated based on the following formula:

Assume that the vector _x0 can be expressed as follows: where the vector _x0 is a vector that lists in order the probability that the target route candidate passes through each point on the route candidate.

のベクトルＲにより表されるものとする。なおベクトルＲは、かかる各評価項目の重要度の比率を順番に並べたベクトルである。評価項目の重要度の比率は、高い値であればその評価項目を優先し、低い値であれば、その評価項目を優先しない指標となる。 In addition, each evaluation item of the route candidate is expressed by the following formula:

The vector R is a vector in which the ratio of the importance of each evaluation item is arranged in order. A high value of the ratio of the importance of an evaluation item is an index for prioritizing that evaluation item, and a low value is an index for not prioritizing that evaluation item.

及び次式

及び次式

及び次式

で与えられ、最適化計算の計算方法として焼きなまし法が指定されているものとする。
ここで、（３）と（４）は、ルートとして必ず選択する必要のある位置を制約するものであり、（５）と（６）は、出力されるルートが０か１となるための制約条件である。 Furthermore, the objective function stored in the optimization parameter database 82 is in the form of a linear function, and the constraint condition is expressed by the following formula:

and the following formula

and the following formula

and the following formula

and the simulated annealing method is specified as the optimization method.
Here, (3) and (4) are constraints on the positions that must be selected as the route, and (5) and (6) are constraints for the output route to be 0 or 1.

のように表すことができ、図１８のステップＳ１２で作成される制約条件は、次式

のように表すことができる。 In this case, the objective function created in step S11 of FIG.

The constraint condition created in step S12 of FIG. 18 can be expressed as follows:

It can be expressed as follows.

のようなルート候補が算出される。 For the objective function and constraint conditions thus expressed, in step S13 of FIG. 18, if the initial point is set to be the same as the route candidate under consideration at that time, the following equation is calculated by the simulated annealing method, for example:

Route candidates such as the following are calculated.

The route candidates calculated at this time utilize the importance ratio of each evaluation item of the route candidates before adjustment, while at the same time satisfying the constraints that must be observed, making it possible to obtain adjusted route candidates with high usability.

As described above, according to the route generation device 80 of this embodiment, the optimization calculation unit 84 generates adjusted route candidates that are adjusted in consideration of specific constraint conditions based on the route candidates generated by the route candidate generation unit 34, so that it is possible to present the user with reasonable route candidates that take such constraint conditions into consideration. Thus, according to this route generation device 80, it is possible to provide the user with an even higher level of satisfaction.

(5) Fifth embodiment In cases where the social impact is high, such as infrastructure, in addition to using the route generation devices 1, 50, 60, and 80 as described above in the first to fourth embodiments, education of users who select routes, etc. is also required. In normal educational settings, education is provided by trainers who are knowledgeable and more experienced, and who evaluate routes generated by trainees (trainees).

However, such educational methods require the time and effort of highly skilled experts. Therefore, in this embodiment, an educational system that utilizes a route generation device (hereinafter referred to as route generation device 1 of the first embodiment) as educational content will be described.

Figure 19 shows the schematic configuration of such an education system 90. This education system 90 is composed of a terminal device 91 used by the trainee, and a route generation system 93 composed of a content generation device 92 and a route generation device 1.

The terminal device 91 is configured as a general-purpose computer device such as a desktop computer device or a laptop computer device, and is connected to the content generating device 92 via a network such as a LAN (Local Area Network) or the Internet (not shown).

As shown in Fig. 20, the content generation device 92 is composed of a general-purpose computer device equipped with a calculation device 100, memory 101, storage device 102, input device 103, and display device 104. The calculation device 100, memory 101, storage device 102, input device 103, and display device 104 have the same configurations and functions as the corresponding parts (calculation device 2, memory 3, storage device 4, input device 5, or display device 6) of the route generation device 1 described above in Fig. 1, so detailed explanations will be omitted here.

The memory 101 of the content generating device 92 stores a content generating program 105, which will be described later. The computing device 100 executes the content generating program 105 to perform various processes of the content generating device 92 as a whole, as will be described later.

Figure 21 shows the flow of processing executed in the education system 90 according to this embodiment. First, in response to a training start request from a user (trainee) using a terminal device 91, the content generation device 92 acquires from the route generation device 1 the performance data D1 (Figure 2) stored in the performance database 20 (Figure 1) of the route generation device 1 (S20).

Then, the content generating device 92 uses the performance data D1 acquired in step S20 to generate a training menu, and displays on the terminal device 91 a practice problem in which the trainee creates a route based on his or her own ideas, similar to a route actually created by an expert in the past, based on the generated training menu (S21).

Then, the content generating device 92 waits for the user to finish generating the route for the exercise (S22), and when the user has finished generating the route, it evaluates the route generated by the trainee based on each past evaluation item (S23), and displays the evaluation result (e.g., similar classes) on the terminal device 91 (S24).Then, the content generating device 92 ends this series of processes.

The upper part of FIG. 22 shows an example of a partial configuration of the exercise screen 110 that the content generating device 92 displays on the terminal device 91 in step S21 of FIG. 21. In this example, one row of the table 111 displayed on the exercise screen 110 corresponds to one exercise, and the difficulty level and the start and end points are displayed for each exercise. When the trainee selects one exercise from these exercises, the content generating device 92 reads out the past performance data D1 corresponding to that exercise from the route generating device 1, and various information required to generate a route contained in the read out performance data D1 is displayed on the terminal device 91. In this way, the trainee refers to this information to generate a route connecting the start and end points.

The lower part of Figure 22 shows an example of a partial configuration of an answer screen 112 that is displayed after the trainee has generated a route for a practice problem. On this answer screen 112, the route generated by the user at that time is displayed as a straight line K1, and a route from the same start point to the end point previously generated by an expert is displayed as a dashed line K2. The answer screen 112 also displays an evaluation of the route generated by the trainee, such as the similarity between these two routes and the cost of actually creating the route created by the trainee, in an evaluation field 113 in the upper right corner of the screen.

According to the educational system 90 of this embodiment as described above, it is possible to provide the user with an evaluation equivalent to that of a highly skilled expert. This is because know-how equivalent to that of an expert is accumulated during the learning phase of the route generation device 1. Furthermore, according to this educational system 90, the importance of each evaluation item can be quantitatively presented, making it possible to pass on know-how in a more objective manner, and it is expected that the trainee's level of proficiency will improve.

(5) Other embodiments In the above-mentioned first to fifth embodiments, the present invention is described as being applied to the route generating devices 1, 50, 60, and 80 that automatically generate routes for infrastructure networks, but the present invention is not limited to this and can be widely applied to other automatic design plan generating devices that automatically generate various other design plans. In this case, it is sufficient to replace "route" in the first to fifth embodiments with the "design plan."

In the above-mentioned first to fifth embodiments, the route generation devices 1, 50, 60, and 80 are constructed using a single computer device. However, the present invention is not limited to this. The route generation devices 1, 50, 60, and 80 may be constructed using multiple computer devices that constitute a distributed computing system, and the functions of the route generation devices 1, 50, 60, and 80 related to the route generation function may be distributed among these multiple computer devices.

Furthermore, in the above-mentioned second to fourth embodiments, the case where the functions according to the second to fourth embodiments are additionally provided to the route generation device 1 according to the first embodiment has been described, but the present invention is not limited to this. For example, the route generation device 50 according to the second embodiment may be equipped with the functions of the route generation device 60 according to the third embodiment or the route generation device 80 according to the fourth embodiment, and furthermore, a route generation device may be constructed by appropriately combining the functions of the route generation devices 1, 50, 60, and 80 according to the first to fourth embodiments.

Similarly, in the above-mentioned fifth embodiment, the route generation device 1 according to the first embodiment is described as being applied as the route generation device, but the present invention is not limited to this, and the route generation devices 50, 60, and 80 according to the second to fourth embodiments may be applied as the route generation device, and further, a route generation device that appropriately combines the functions of the route generation devices 1, 50, 60, and 80 according to the first to fourth embodiments may be applied.

Furthermore, in the above-mentioned fifth embodiment, the route generation system 93 is described as being composed of the content generation device 92 and the route generation device 1, but the present invention is not limited to this. All the functions of the content generation device 92 may be installed in the route generation device 1, and the route generation system 93 may be composed of only the route generation device 1.

The present invention relates to a design plan generation device, and can be applied, for example, to a route generation device that automatically generates routes for infrastructure networks.

1, 50, 80...Route generation device, 2...Calculation device, 5...Input device, 6...Display device, 10...Data characteristic classification program, 11...Data characteristic learning program, 12...Evaluation item calculation program, 13...Route learning program, 14...Route candidate generation program, 15, 61...Result display program, 20...Performance database, 21...Learning database, 22...Plan database, 23...Route candidate database, 24...Route evaluation item database, 30...Data characteristic classification unit, 31...Data characteristic learning unit, 32...Evaluation item calculation unit, 33... Route learning unit, 34...route candidate generation unit, 35, 63...result display unit, 40, 70...result display screen, 42...route map, 43...graph, 51...data filtering program, 52...data filtering unit, 62...result evaluation acquisition program, 64...result evaluation acquisition unit, 81...optimization calculation program, 82...optimization parameter database, 83...adjusted route candidate database, 84...optimization calculation unit, 90...education system, 92...content generation device, 105...content generation program, D1...performance data, D2...planning data.

Claims (10)

A design plan generation system for generating a new design plan based on a past design plan,
A classification unit that classifies the past design plans into one or more classes based on the past design plans and performance data including various information required for formulating the past design plans;
a characteristic learning unit that learns a class classification according to data characteristics based on the performance data of the past design plans that belong to each of the classes classified by the classification unit;
an evaluation item calculation unit that calculates, for each of the classes, information that is considered important among the various information included in the performance data based on a learning result of the characteristic learning unit, as an evaluation item for the class;
a design plan learning unit configured to learn, for each of the classes, the past design plans belonging to the class;
a design plan generating unit that generates the new design plan for each of the classes based on the evaluation items for each of the classes calculated by the evaluation item calculating unit, a learning result of the past design plan for each of the classes by the design plan learning unit, and various information required for planning the new design plan, and calculates the evaluation items in the new design plan for each of the classes;
a result display unit that displays the new design plan for each class generated by the design plan generation unit and the evaluation items in the new design plan;
Equipped with
a characteristic learning unit that learns classifying the past design plans according to data characteristics based on a user's evaluation of the evaluation items for each class calculated by the evaluation item calculation unit and the learning results of the past design plans for each class by the design plan learning unit, the characteristic learning unit that learns class classification according to data characteristics, the evaluation item calculation unit that calculates the evaluation items for each class, and the design plan learning unit that learns the past design plans. A design plan generation system for generating a new design plan based on a past design plan,
A classification unit that classifies the past design plans into one or more classes based on the past design plans and performance data including various information required for formulating the past design plans;
a characteristic learning unit that learns a class classification according to data characteristics based on the performance data of the past design plans that belong to each of the classes classified by the classification unit;
an evaluation item calculation unit that calculates, for each of the classes, information that is considered important among the various information included in the performance data based on a learning result of the characteristic learning unit, as an evaluation item for the class;
a design plan learning unit configured to learn, for each of the classes, the past design plans belonging to the class;
a design plan generating unit that generates the new design plan for each of the classes based on the evaluation items for each of the classes calculated by the evaluation item calculating unit, a learning result of the past design plan for each of the classes by the design plan learning unit, and various information required for planning the new design plan, and calculates the evaluation items in the new design plan for each of the classes;
a result display unit that displays the new design plan for each class generated by the design plan generation unit and the evaluation items in the new design plan;
a filtering unit that identifies bad data that satisfies a preset condition from among the past design plans and executes a filtering process to remove the identified bad data;
Equipped with
The classification unit includes:
classifying the past design plans from which the bad data has been removed by the filtering process into one or more classes. A design plan generation system for generating a new design plan based on a past design plan,
A classification unit that classifies the past design plans into one or more classes based on the past design plans and performance data including various information required for formulating the past design plans;
a characteristic learning unit that learns a class classification according to data characteristics based on the performance data of the past design plans that belong to each of the classes classified by the classification unit;
an evaluation item calculation unit that calculates, for each of the classes, information that is considered important among the various information included in the performance data based on a learning result of the characteristic learning unit, as an evaluation item for the class;
a design plan learning unit configured to learn, for each of the classes, the past design plans belonging to the class;
a design plan generating unit that generates the new design plan for each of the classes based on the evaluation items for each of the classes calculated by the evaluation item calculating unit, a learning result of the past design plan for each of the classes by the design plan learning unit, and various information required for planning the new design plan, and calculates the evaluation items in the new design plan for each of the classes;
a result display unit that displays the new design plan for each class generated by the design plan generation unit and the evaluation items in the new design plan;
Equipped with
a design plan generating system, characterized in that, based on the new design plan displayed by the result display unit and a user's evaluation result for the evaluation items in the new design plan, the classification unit re-executes classification of the past design plan, the characteristic learning unit re-executes learning of class classification according to data characteristics, the evaluation item calculation unit re- executes calculation of the evaluation items for each class, and the design plan learning unit re-executes learning of the past design plan. A design plan generation system for generating a new design plan based on a past design plan,
A classification unit that classifies the past design plans into one or more classes based on the past design plans and performance data including various information required for formulating the past design plans;
a characteristic learning unit that learns a class classification according to data characteristics based on the performance data of the past design plans that belong to each of the classes classified by the classification unit;
an evaluation item calculation unit that calculates, for each of the classes, information that is considered important among the various information included in the performance data based on a learning result of the characteristic learning unit, as an evaluation item for the class;
a design plan learning unit configured to learn, for each of the classes, the past design plans belonging to the class;
a design plan generating unit that generates the new design plan for each of the classes based on the evaluation items for each of the classes calculated by the evaluation item calculating unit, a learning result of the past design plan for each of the classes by the design plan learning unit, and various information required for planning the new design plan, and calculates the evaluation items in the new design plan for each of the classes;
a result display unit that displays the new design plan for each class generated by the design plan generation unit and the evaluation items in the new design plan;
an optimization calculation unit that calculates a design plan and an evaluation item that satisfy a constraint condition specified by a user through optimization calculation processing based on the new design plan generated or calculated by the design plan generation unit and the evaluation item in the new design plan;
A design plan generating system comprising: A design plan generation system for generating a new design plan based on a past design plan,
A classification unit that classifies the past design plans into one or more classes based on the past design plans and performance data including various information required for formulating the past design plans;
a characteristic learning unit that learns a class classification according to data characteristics based on the performance data of the past design plans that belong to each of the classes classified by the classification unit;
an evaluation item calculation unit that calculates, for each of the classes, information that is considered important among the various information included in the performance data based on a learning result of the characteristic learning unit, as an evaluation item for the class;
a design plan learning unit configured to learn, for each of the classes, the past design plans belonging to the class;
a design plan generating unit that generates the new design plan for each of the classes based on the evaluation items for each of the classes calculated by the evaluation item calculating unit, a learning result of the past design plan for each of the classes by the design plan learning unit, and various information required for planning the new design plan, and calculates the evaluation items in the new design plan for each of the classes;
a result display unit that displays the new design plan for each class generated by the design plan generation unit and the evaluation items in the new design plan;
a content generating unit that generates exercises for a user to generate a design plan based on performance data of the past design plans, evaluates the design plan created by the user for the generated exercises based on the past design plans, and presents the evaluation results to the user;
A design plan generating system comprising : A design plan generation method executed in a design plan generation system that generates a new design plan based on a past design plan, comprising:
A first step of classifying the past design plans into one or more classes based on the past design plans and performance data including various information required for formulating the past design plans;
a second step of learning a class classification according to data characteristics based on the performance data of the past design plans belonging to each of the classified classes, and calculating, for each of the classes, information that is considered important among the various information included in the performance data as an evaluation item for the class based on a learning result, and learning, for each of the classes, the past design plans belonging to the class;
a third step of generating the new design plan for each of the classes based on the calculated evaluation items for each of the classes, learning results of the past design plans for each of the classes, and various information required for planning the new design plan, and calculating the evaluation items in the new design plan for each of the classes;
a fourth step of displaying the generated new design plan and the evaluation items in the new design plan;
Equipped with
and re-executing the processes of the first and second steps based on a user's evaluation of the calculated evaluation items for each class and learning results of the past design plans for each class. A design plan generation method executed in a design plan generation system that generates a new design plan based on a past design plan, comprising:
A first step of classifying the past design plans into one or more classes based on the past design plans and performance data including various information required for formulating the past design plans;
a second step of learning a class classification according to data characteristics based on the performance data of the past design plans belonging to each of the classified classes, and calculating, for each of the classes, information that is considered important among the various information included in the performance data as an evaluation item for the class based on a learning result, and learning, for each of the classes, the past design plans belonging to the class;
a third step of generating the new design plan for each of the classes based on the calculated evaluation items for each of the classes, learning results of the past design plans for each of the classes, and various information required for planning the new design plan, and calculating the evaluation items in the new design plan for each of the classes;
a fourth step of displaying the generated new design plan and the evaluation items in the new design plan;
Equipped with
In the first step,
identifying bad data that satisfies a preset condition from the past design plans, performing a filtering process to remove the identified bad data, and classifying the filtered past design plans into one or more classes. A design plan generation method executed in a design plan generation system that generates a new design plan based on a past design plan, comprising:
A first step of classifying the past design plans into one or more classes based on the past design plans and performance data including various information required for formulating the past design plans;
a second step of learning a class classification according to data characteristics based on the performance data of the past design plans belonging to each of the classified classes, and calculating, for each of the classes, information that is considered important among the various information included in the performance data as an evaluation item for the class based on a learning result, and learning, for each of the classes, the past design plans belonging to the class;
a third step of generating the new design plan for each of the classes based on the calculated evaluation items for each of the classes, learning results of the past design plans for each of the classes, and various information required for planning the new design plan, and calculating the evaluation items in the new design plan for each of the classes;
a fourth step of displaying the generated new design plan and the evaluation items in the new design plan;
Equipped with
and re-executing the processes of the first and second steps based on the displayed new design plan and a user's evaluation result for the evaluation items in the new design plan. A design plan generation method executed in a design plan generation system that generates a new design plan based on a past design plan, comprising:
A first step of classifying the past design plans into one or more classes based on the past design plans and performance data including various information required for formulating the past design plans;
a second step of learning a class classification according to data characteristics based on the performance data of the past design plans belonging to each of the classified classes, and calculating, for each of the classes, information that is considered important among the various information included in the performance data as an evaluation item for the class based on a learning result, and learning, for each of the classes, the past design plans belonging to the class;
a third step of generating the new design plan for each of the classes based on the calculated evaluation items for each of the classes, learning results of the past design plans for each of the classes, and various information required for planning the new design plan, and calculating the evaluation items in the new design plan for each of the classes;
a fourth step of displaying the generated new design plan and the evaluation items in the new design plan;
Equipped with
In the third step,
calculating, by optimization calculation processing, a design plan and evaluation items that satisfy constraint conditions specified by a user, based on the generated or calculated new design plan and the evaluation items in the new design plan. A design plan generation method executed in a design plan generation system that generates a new design plan based on a past design plan, comprising:
A first step of classifying the past design plans into one or more classes based on the past design plans and performance data including various information required for formulating the past design plans;
a second step of learning a class classification according to data characteristics based on the performance data of the past design plans belonging to each of the classified classes, and calculating, for each of the classes, information that is considered important among the various information included in the performance data as an evaluation item for the class based on a learning result, and learning, for each of the classes, the past design plans belonging to the class;
a third step of generating the new design plan for each of the classes based on the calculated evaluation items for each of the classes, learning results of the past design plans for each of the classes, and various information required for planning the new design plan, and calculating the evaluation items in the new design plan for each of the classes;
A fourth step of displaying the generated new design plan and the evaluation items in the new design plan;
a fifth step of generating exercises for a user to generate a design plan based on performance data of the past design plans, evaluating the design plan created by the user for the generated exercises based on the past design plans, and presenting the evaluation results to the user ;
A design plan generating method comprising :

Images