JP7823908B2 - Setting support device, setting support system, setting support method, and program - Google Patents

Description

The present invention relates to a configuration support device, configuration support system, configuration support method, and program that can easily solve configuration problems.

For devices that require user configuration, there is a large gap between the problem the user has, such as poor signal reception on a smartphone, and the solution, such as changing specific wireless communication parameters on the device. The reason for this gap is that users do not know which setting parameters they need to select and change to solve the device's problem. On the other hand, configuring devices to solve their own problems is impossible without taking into account the settings within each device.

Patent Document 1 discloses the following estimation device, etc. The specification of the invention states, "The estimation unit 11 functions as estimation means for estimating an evaluation score (the relative value of the communication quality after the base station parameter change to the communication quality before the base station parameter change) indicating the degree of improvement in communication quality when the base station parameters of the estimation target base station are changed from those before the change to those after the change, based on the communication quality measurement information and base station parameter setting information of the estimation target base station acquired by the information acquisition unit 13. Specifically, the estimation unit 11 receives as input the communication quality measurement information and base station parameter setting information of the estimation target base station before the change, and calculates, for each of the six types of base station parameter sets Set0 to SetE described below, each evaluation score (the relative value of the communication quality after the base station parameter change to the communication quality before the change) after the change to the base station parameter set."

Japanese Patent Application Laid-Open No. 2020-178266

The disclosures of the above prior art documents are incorporated herein by reference. The following analysis was conducted by the inventors.

As described above, in the invention disclosed in Patent Document 1, the estimation unit estimates an evaluation score that indicates the degree of improvement in communication quality after changing a base station parameter set, which is a combination of base station parameters, based on communication quality measurement information and base station parameter setting information, thereby identifying a base station parameter set from among multiple parameter sets that will improve communication quality. In other words, it is possible to identify parameters whose settings should be changed.

However, the invention disclosed in Patent Document 1 does not accept problems with each device (base station), and does not provide a problem-solving method that allows users to easily input a problem and output setting changes that address that problem.

Therefore, one aspect of the present invention is to provide a configuration support device, configuration support system, configuration support method, and program that allows users to easily solve configuration problems.

According to a first aspect of the present invention, there is provided a configuration assistance device having a configuration problem acquisition unit that acquires a configuration problem related to device configuration, a device parameter acquisition unit that acquires device parameters including the device configuration, a calculation unit that calculates corresponding parameters corresponding to the configuration problem based on the configuration problem and the device parameters, and an output unit that outputs the corresponding parameters.

In a second aspect of the present invention, there is provided a setting support system including a setting device and a setting support server device, wherein the setting device has a setting problem acquisition unit that acquires a setting problem related to configuration, a device parameter acquisition unit that acquires device parameters including the settings of the setting device, and an input data transmission unit that transmits the setting problem and the device parameters, the setting support server device has a calculation unit that uses a learning model to calculate corresponding parameters corresponding to the setting problem based on the setting problem and the device parameters, and a corresponding parameter transmission unit that transmits the corresponding parameters, and the setting device further has an output unit that outputs the corresponding parameters.

In a third aspect, the present invention provides a configuration assistance method that executes the following steps by a computer, the configuration assistance method including the steps of: acquiring a configuration problem that is a problem related to device configuration; acquiring device parameters including the device configuration; calculating corresponding parameters corresponding to the configuration problem based on the configuration problem and the device parameters using a learning model; and outputting the corresponding parameters.

In a fourth aspect of the present invention, a program is provided for causing a computer to execute the following steps: acquiring a configuration problem, which is a problem related to device settings; acquiring device parameters including the device settings; calculating corresponding parameters corresponding to the configuration problem using a learning model based on the configuration problem and the device parameters; and outputting the corresponding parameters.

This program can be recorded on a computer-readable storage medium. The storage medium can be a non-transient medium such as a semiconductor memory, hard disk, magnetic recording medium, or optical recording medium. The present invention can also be embodied as a computer program product.

Each aspect of the present invention makes it possible to provide a configuration assistance device, configuration assistance system, configuration assistance method, and program that enable users to easily solve configuration problems.

1 is a block diagram illustrating an example of a configuration of a setting support device according to the present disclosure. FIG. 1 is a diagram illustrating an overview of processing in a setting support device according to the present disclosure. 1 is a block diagram illustrating an example of a configuration of a setting support device according to the present disclosure. 10 is a flowchart illustrating an operation of the setting support device of the present disclosure. 1 is a block diagram illustrating an example of a hardware configuration of a setting assistance device according to the present disclosure. 1 is a block diagram illustrating an example of a configuration of a setting support device according to the present disclosure. 10 is a flowchart illustrating an operation of the setting support device of the present disclosure. FIG. 1 is a diagram illustrating an overview of processing of a setting support system according to the present disclosure. 10 is a flowchart illustrating the operation of the setting support system of the present disclosure. 10 is a flowchart for explaining the operation of the learning model update process after outputting corresponding parameters in the setting support system of the present disclosure. FIG. 1 is a diagram illustrating an overview of an embodiment of a setting support system according to the present disclosure. FIG. 1 illustrates an example of the configuration of an embodiment of a setting support system according to the present disclosure. FIG. 10 is a diagram illustrating a learning data set in an embodiment of the setting support system of the present disclosure. FIG. 10 is a sequence diagram for explaining the operation of an embodiment of the setting support system of the present disclosure. 10 is a flowchart illustrating an operation during continuous learning of an embodiment of the setting support system of the present disclosure. FIG. 10 is a diagram illustrating a configuration in which the configuration support system according to an embodiment of the present disclosure is completed within a wireless device. FIG. 10 is a diagram illustrating a configuration in which only continuous learning is performed outside the wireless device in an embodiment of the setting support system of the present disclosure.

First, an overview of one embodiment will be described. Note that the drawing reference numbers in this overview are added to each element as an example for convenience to aid understanding, and the description of this overview is not intended to be limiting in any way. In this disclosure, the drawings relate to one or more embodiments.

Figure 1 shows a block diagram illustrating an example of the configuration of the setting assistance device disclosed herein. The setting assistance device 10 disclosed herein includes a setting problem acquisition unit 11, a device parameter acquisition unit 12, a calculation unit 13, and an output unit 14.

The setting problem acquisition unit 11 acquires a setting problem, which is a problem related to device settings. The device parameter acquisition unit 12 acquires device parameters including the device settings. The calculation unit 13 uses a learning model to calculate corresponding parameters corresponding to the setting problem based on the setting problem and the device parameters, and the output unit outputs the corresponding parameters.

In this way, the configuration assistance device disclosed herein accepts from the user configuration problems with the device for which configuration assistance is being provided (for example, in the case of a communications device, audio cuts out, noise is present, etc.), and then obtains the device parameters, which are the current configuration parameters of that device. Using a learning model based on the configuration problem and the device parameters, it is possible to calculate and output corresponding parameters, which are configuration parameters that correspond to the configuration problem. Therefore, even users with little knowledge about device configuration, such as end users, can input a configuration problem into the device and learn corresponding parameters that will assist in resolving the configuration problem.

[First embodiment]
[Processing Overview]
FIG. 2 is a diagram illustrating an overview of the processing in the setting assistance device of the present disclosure. As shown in this diagram, there is a setting assistance device 10, a user 1, and a device 2 to be assisted (hereinafter referred to as the target device 2). The target device 2 and the setting assistance device 10 may be integrated or separate devices. User 1 is the user of the target device 2. The setting assistance device 10 acquires a setting problem for the target device 2 from the user 1 in natural language. The setting assistance device 10 also acquires device setting parameters Sp related to the setting problem from the target device 2. The setting assistance device 10 uses learning models 3-5 to calculate corresponding parameters Sa corresponding to the acquired setting problem. The number of elements of the acquired setting parameters Sp and the calculated corresponding parameters Sa may differ (i.e., n ≠ m in FIG. 2).

Learning model 3-5 uses corresponding parameters as training signals in advance, and learns by inputting a configuration problem in natural language and the device configuration parameters related to that configuration problem. Note that the example in Figure 2 is a schematic representation of a neural network model, and the number of elements in the model (shown by circles) and the form of connections between elements (shown by wires between elements) are not limited to those shown in Figure 2.

[Device Configuration]
3 is a block diagram showing an example of the configuration of the setting support device 10 of the present disclosure. As shown in this figure, the setting support device 10 of the present disclosure includes a setting problem acquisition unit 11, a device parameter acquisition unit 12, a calculation unit 13, an output unit 14, and a learning model generation unit 15.

The configuration problem acquisition unit 11 acquires configuration problems related to the configuration of the device. In the case of a communications device, a "configuration problem" is a sentence, word, phrase, or the like that indicates an inconvenient event for the user, such as "audio cuts out," "audio is distorted," or "noise is present." The configuration assistance device 10 disclosed herein executes the following processing to resolve this configuration problem, and calculates and outputs corresponding parameters as a means of resolution. As mentioned above, this unit may acquire configuration problems in natural language.

The hypothesis problem acquisition unit 11 may also convert the hypothesis problems written in natural language into vector expressions, allowing for similarity between the hypothesis problems. To generate vector expressions, Word2Vec or Doc2Vec, which can convert sentences of any length into vectors of a fixed length, can be used.

The device parameter acquisition unit 12 acquires device parameters including device settings. In Figure 2, it acquires device parameters that indicate the current settings of the target device 2. In other words, it acquires setting values and converts them appropriately into a form suitable for calculation by the calculation unit 13, which will be described later. Specifically, the actual measured values are normalized, for example, and categorical data is converted using a quantification method such as one-hot encoding.

The calculation unit 13 uses a learning model to calculate corresponding parameters corresponding to the set problem based on the set problem and the device parameters. The "learning model" is a model for calculating corresponding parameters learned by the learning model generation unit 15 described below. Calculation involves inputting the set problem and the device parameter values into the learning model and obtaining the corresponding parameter values that are the calculation results of the learning model.

As shown in Figure 2, user 1 inputs using natural language, and target device 2 inputs the setting parameters Sp and calculates the corresponding parameters Sa using learning model 3-5.

The output unit 14 outputs the corresponding parameters. It performs processing such as displaying the parameters on a display device or transmitting the parameters to other devices via communication.

The learning model generation unit 15 uses the setting problem and device parameters as input data and generates a learning model trained using learning data in which the setting problem and corresponding parameters corresponding to the device parameters are used as training data. Learning models 3-5 in Figure 2 are shown in a schematic diagram with neural networks in mind, but other supervised or unsupervised models can also be used. While Figure 2 illustrates three models, models 3-5, these may also be integrated into a single model. That is, models 3-5 may be combined from the beginning and training may proceed, or models 3 and 4 may be trained first, and then model 5 may be combined and training may proceed.

For example, during learning, unsupervised learning can be performed by vectorizing the text input as learning data using the aforementioned Doc2Vec or the vector representation of the device parameters and reducing the dimension of the vector using a dimension reduction algorithm such as principal component analysis. In this case, a principal component with a predetermined cumulative contribution rate can be selected, and the dimension-reduced learning data can be output and used as input values for Model 5 to train Model 5 again.

In addition, the learning may be performed by providing the parameter values themselves as training data, or by providing information on which parameters are related. In other words, the vector of corresponding parameters Sa in Figure 2, which is the calculation result of the calculation unit 13, may take on a value of 0 or 1. In this case, the user can know which parameters to change as setting items, rather than the parameter values.

[Operation description]
4 is a flowchart illustrating the operation of the configuration assistance device 10 of the present disclosure. As shown in this figure, when the configuration assistance device 10 starts operation, it accepts and acquires a configuration problem from the user (step S41). Next, it acquires device parameters, including device settings, from the target device (step S42). Based on the configuration problem and the device parameters, it calculates corresponding parameters corresponding to the configuration problem using a learning model (step S43). Finally, it outputs the calculated corresponding parameters (step S44).

[Hardware configuration]
Next, a description will be given of the hardware configuration of the setting assistance device according to the first embodiment. Fig. 5 is a block diagram showing an example of the hardware configuration of the setting assistance device 10 according to the first embodiment.

The setting assistance device 10 can be configured as an information processing device (computer) and has the configuration shown in FIG. 5. For example, the setting assistance device 10 includes a CPU (Central Processing Unit) 51, memory 52, an input/output interface 53, and a NIC (Network Interface Card) 54, which serves as a communication means, all of which are interconnected by an internal bus 55.

However, the configuration shown in Figure 5 is not intended to limit the hardware configuration of the devices that make up the setting assistance device 10. Each setting assistance device 10 may include hardware not shown, and may not include an input/output interface 53 if necessary. Furthermore, the number of CPUs and other components included in the setting assistance device 10 is not intended to be limited to the example shown in Figure 5; for example, each device may include multiple CPUs.

Memory 52 is RAM (Random Access Memory), ROM (Read Only Memory), and auxiliary storage device (hard disk, etc.).

The input/output interface 53 is a means for interfacing with a display device and an input device (not shown). The display device is, for example, an LCD display. The input device is, for example, a device that accepts user operations, such as a keyboard or mouse.

When calculating corresponding parameters, the functions of the setting assistance device 10 are realized by processing modules: a setting problem acquisition program, a device parameter acquisition program, a calculation program, an output program, and a set of parameters for a trained learning model stored in memory 52, etc. When generating a learning model, the functions are realized by a learning model generation program and a dataset consisting of input learning data and training data.

The above processing modules are implemented, for example, by the CPU 51 executing programs stored in the memory 52. Furthermore, the programs can be updated by downloading them over a network or using a storage medium on which the programs are stored. The above processing modules may also be implemented using semiconductor chips. In other words, it is sufficient to have some means of executing the functions performed by the above processing modules using some kind of hardware and/or software.

[Hardware operation: learning]
First, a learning model is generated in the setting assistance device 10. A learning model generation program is called from the memory 52 and executed by the CPU 51. The program sequentially reads learning data sets stored in the memory 52 and adjusts the parameters of the learning model using the input learning data and teacher data. The parameters may be adjusted using a loss function such as RSS (Residual Sum of Squares), an existing technology. If the model is a neural network model with multiple layers, a method such as error propagation may be used.

[Hardware operation: calculation time]
After learning is complete, in the setting assistance device 10, a setting problem acquisition program is called from the memory 52 and executed by the CPU 51. This program acquires the setting problems via the input/output interface 53 and the NIC 54. If the setting problems are voice, a voice recognition process is executed here to convert the voice into text. The text setting problems are vectorized using a program using existing technology and temporarily stored in the memory 52. Next, a device parameter acquisition program is called from the memory 52 and executed by the CPU 51. This program acquires device parameters, which are setting parameters for the target device. Specifically, this program may send a request to the target device via the NIC 54 and acquire the parameters in response. The acquired parameters are temporarily stored in the memory 52.

Next, the calculation program is called from memory 52 and executed by CPU 51. This program reads the trained learning model stored in memory 52. Next, the program executes calculations using the set problem (vector) and device parameters (vector) temporarily stored in memory 52 as input. At this time, the input data is subjected to standardization and other processes as appropriate. As a result of the calculations, the corresponding parameters (vector) are calculated. Next, the output program is called from memory 52 and executed by CPU 51. This program outputs the corresponding parameters via NIC 54 and input/output interface 53.

[Effect description]
As described above, the setting assistance device 10 of the present disclosure can acquire setting problems that a user has with a target device, and calculate corresponding parameters for resolving the setting problems based on the setting problems and the device parameters of the target device. Therefore, it is possible to provide a setting assistance device, setting assistance system, setting assistance method, and program that enable a user to easily perform optimal settings based on the corresponding parameters output.

Second Embodiment
In the second embodiment, based on the first embodiment, a setting support device is provided that can further update an already learned learning model based on the determination result of whether or not the corresponding parameters have been reflected in the target device.

[Device Configuration]
6 is a block diagram showing an example of the configuration of the setting support device 10 of the present disclosure. The setting support device 10 of the present disclosure includes a setting problem acquisition unit 11, a device parameter acquisition unit 12, a calculation unit 13, an output unit 14, and a learning model generation unit 15. These components have already been described in the above embodiment, so description thereof will be omitted. The setting support device 10 in this embodiment is characterized in that it further includes a determination unit 16, and that the learning model generation unit updates the learning model in accordance with the determination result of the determination unit 16.

The judgment unit 16 judges whether the corresponding parameters have been reflected in the device. The corresponding parameters calculated by the learning model and output by the output unit 14 can be changed by user operation. Here, the judgment unit 16 compares the calculated and output corresponding parameters with the latest device parameters acquired by the device parameter acquisition unit 12, and judges whether they match in part or in whole. If the judgment results in a match, it considers that the corresponding parameters have been reflected in the device parameters of the target device, and sends a trigger to update the learning model to the learning model generation unit 15.

The learning model generation unit 15 updates the learning model according to the judgment result of the judgment unit 16. Specifically, for example, if it is judged as a result of the judgment that the corresponding parameter has been reflected as a device parameter in the target device, the learning model may be updated based on the corresponding parameter, the setting problem that is the input data from which the corresponding parameter was calculated, and the device parameters. Specifically, so-called online learning may be performed using, for example, the setting problem, the device parameters, and the calculated corresponding parameter. Here, if the weight between elements of the learning model is Wn, the value of the loss function is l, and the learning rate is α, then

In the above example, online learning is performed each time a corresponding parameter is output, and the learning model is updated. However, multiple outputs may be collected as a learning dataset, and mini-batch learning may be performed.

[Operation description]
FIG. 7 is a flowchart illustrating the operation of the setting assistance device 10 of the present disclosure. The process from the start of operation to the output of the corresponding parameters (steps S71 to S74, not shown) has already been described in FIG. 4 (steps S41 to S44), so a description thereof will be omitted. Once the corresponding parameters are output, it is determined whether the corresponding parameters have been reflected in the device (step S75). If it is determined that the corresponding parameters have been reflected (step S75, Y), the learning model is updated based on the corresponding parameters determined to have been reflected in the device, the setting problem that is the input data from which the corresponding parameters were calculated, and the device parameters (step S76). If it is determined that the corresponding parameters have not been reflected (step S75, N), the series of processes ends.

[Effect description]
The setting assistance device 10 of the present disclosure can update the learning model based on whether the output corresponding parameters are reflected in the device parameters. This makes it possible to generate a learning model that is more optimized for the setting problem, and to output more appropriate corresponding parameters to the user.

[Third embodiment]
In the third embodiment, a setting support system including a setting device 20 and a setting support server device 30 will be described.

[Processing configuration]
FIG. 8 is a diagram illustrating an overview of the processing of the setting support system of the present disclosure. As shown in this diagram, the setting support system includes a setting device 20 and a setting support server device 30. The device to be set and the setting device 20 may be integrated or separate devices. The setting device 20 includes a setting question acquisition unit 11, a device parameter acquisition unit 12, an output unit 14, a determination unit 16, and an input data transmission unit 17. The setting support server device 30 includes a calculation unit 13, a learning model generation unit 15, and a corresponding parameter transmission unit 18. The setting support system of the present disclosure is characterized in that the setting device 20 includes the input data transmission unit 17, and the setting support server device 30 includes the corresponding parameter transmission unit 18. Other components have been described in the above embodiment, so description thereof will be omitted.

Here, in the configuration support system disclosed herein, multiple configuration devices may be connected to a single configuration support server device 30 and share learning models.

The input data transmission unit 17 transmits the setting questions and device parameters acquired by the setting device 20 to the setting assistance server device 30.

The corresponding parameter transmission unit 18 transmits the corresponding parameters calculated by the setting support server device 30 to the setting device 20.

[Operation description]
9 is a flowchart for explaining the operation of the setting support system of the present disclosure. As shown in this figure, the setting device 20 acquires a setting problem (step S91). Next, device parameters are acquired (step S92). The acquired setting problem and device parameters are transmitted (step S93). Thereafter, the setting support server device 30 uses a learning model to calculate corresponding parameters corresponding to the setting problem based on the received setting problem and the device parameters (step S94). The calculated corresponding parameters are transmitted (step S95). The setting device 20 receives the corresponding parameters and outputs them (step S96).

Figure 10 is a flowchart illustrating the operation of the learning model update process after the corresponding parameters are output in the configuration support system of the present disclosure. As shown in this figure, after the corresponding parameters are output in the configuration device 20, it is determined whether the corresponding parameters have been reflected in the device (step S1017). If it is determined that the corresponding parameters have been reflected (step S1017, Y), the configuration support server device 30 updates the learning model based on the corresponding parameters determined to have been reflected in the device, the configuration problem that is the input data from which the corresponding parameters were calculated, and the device parameters (step S1018). If it is determined that the corresponding parameters have not been reflected (step 1017, N), the series of processes ends.

[Effect description]
According to the setting support system of the present disclosure, it is possible to calculate corresponding parameters corresponding to setting problems in cooperation with both the setting device 20 and the setting support server device 30. Since the calculation process of corresponding parameters is centralized in the setting support server device 30, even when the processing power of the setting device 20 is limited, such as in a small mobile terminal, the setting device 20 can make an inquiry to the setting support server device, allowing for rapid processing.

In addition, multiple setting devices 20 can be connected to the setting assistance server device 30, and by generating and updating learning models using multiple setting devices 20 and the setting assistance server device 30, it is possible to provide learning models with greater versatility.

[Example]
An embodiment of the setting support system of the present disclosure will be described below. The following is an embodiment of the setting support system in which the setting device is a wireless device.

[Outline of the Example]
FIG. 11 is a diagram illustrating an overview of an embodiment of a configuration support system according to the present disclosure. As shown in this diagram, a user inquires of a wireless device (corresponding to a configuration device) that the line is unstable. At this time, band steering is set to OFF as a device parameter in the wireless device. The wireless device sends the user's inquiry (wireless quality problem) and the device parameters to an analysis server. The analysis server classifies the query using a machine learning model and returns the optimal setting result to the wireless device. The wireless device presents the optimal setting, with band steering set to ON, to the user.

[Configuration of the Example]
12 is a diagram showing an example of the configuration of an embodiment of a setting support system according to the present disclosure. As shown in this diagram, the setting support function of the setting support system according to the present invention is mainly realized by the wireless device 101 and a machine learning model in the analysis server 201. Depending on the capacity of the machine learning model, it may be possible to operate it by incorporating it into the wireless device 101.

The wireless device 101 has six functions related to the invention, numbered 102 to 107. The natural language input function 102 accepts the user's wireless quality problems in natural language format and transmits them to the analysis server for classification. Note that if the user's input is text input using a screen, it is transmitted directly to the analysis server 201, and if the input is voice, the voice waveform is converted into text format and then transmitted in text format to the analysis server 201.

The device parameter acquisition function 103 acquires wireless device parameters (radio signal strength, bandwidth, interference information, device settings, etc.). This acquired data is also sent to the analysis server 201 for classification.

The setting presentation function 104 presents the user with the optimal settings classified by the machine learning model and the wireless quality problems they solve.

The setting selection/completion input function 105 allows the user to select whether to reflect the presented settings on the device, and if necessary, allows the user to enter a small amount of complementary information (such as a password).

The setting reflection function 106 reflects the optimal settings classified based on the received supplementary information.

The wireless communication quality acquisition function 107 has the function of acquiring the current wireless communication quality (communication stability, delay, packet loss, etc.) and the wireless quality before the settings were applied from past communication quality logs. The acquired wireless communication quality is then sent to the continuous learning function 207 of the analysis server 201.

The functions related to the invention within the analysis server 201 are mainly the machine learning model 202 and the continuous learning function 207.

The internal structure of the machine learning model 202 is divided into four components, 203 to 206. The natural language processing module 203 accepts users' wireless quality problems in natural language format, interprets their meaning, and encodes them into a format that is easy for machines to use. The device parameter encoding module 204 encodes device parameters into a format that is easy for machines to use, such as numbers or vectors (one-hot encoding for categories, scaling for numbers, etc.). Note that it is necessary to decide in advance which device parameters to use. The representation integration module 205 inputs the outputs of the natural language processing module 203 and the device parameter encoding module 204 and integrates the two features.

The classification module 206 inputs the outputs of the integration module 205 for both representations and performs classification. Furthermore, multiple classifications are performed by applying a softmax function to the fully connected layer. The number of classification targets must be determined in advance during the learning stage, and the format of the objective variable (O1 in Figure 13) must also match this. Note that when the input and output of the machine learning model are referred to in this description, this refers to the input and output of the machine learning model 202.

The continuous learning function 207 provides feedback on the user's use of the present invention and uses it for continuous learning.

Figure 13 shows the learning dataset for the configuration support system of this disclosure. As shown in this figure, supervised learning is performed using two explanatory variables: the user's wireless quality problem E1 and the device parameter E2, and the optimal wireless device setting O1 as the objective variable. The following describes their structure and correlation.

The user's wireless quality problem E1 is part of the input data for the machine learning model of this invention, and indicates the problems and improvement requests the user has regarding wireless communication quality. These are accepted in natural language format and input into the model as explanatory variables. The role of these variables is to enable the model to classify and suggest settings that correspond to the user's specific quality improvement requests.

Device parameters E2 represent information about the wireless device (such as signal strength, bandwidth, and device settings). These parameters are acquired by the wireless device and used as input data for the machine learning model. The role of device parameters E2 is to enable the model to classify and suggest appropriate settings taking into account the current state of the wireless device. This allows the model to present optimal settings tailored to each individual user's environment.

The optimal wireless device settings O1 are the objective variables of the machine learning model and represent the optimal settings for the wireless device to be selected based on the user's wireless quality problem E1 and device parameters E2. The machine learning model classifies and predicts the optimal settings from these explanatory variables and presents them to the user. Continuous learning is performed taking into account whether the user has reflected the presented settings and any changes in wireless quality due to the reflection of the settings, but the basic configuration is the same.

Since there is a correlation between the explanatory variables and the objective variable, this will be explained.
Regarding the correlation between the user's wireless quality problem E1 and the optimal wireless device setting O1, the wireless quality problem the user is experiencing (such as a decrease in communication speed or a decrease in connection stability) directly affects the selection of the optimal wireless device setting. Since a setting is required to address a specific problem, it is considered that there is a correlation between the user's wireless quality problem E1 and the optimal setting O1.

Regarding the correlation between device parameters E2 and optimal wireless device settings O1, device parameters that reflect the current state of the wireless device (such as signal strength, bandwidth, and current settings) also influence the determination of optimal wireless device settings. Because it is necessary to select appropriate settings that adapt to the environment, it is believed that a correlation also exists between device parameters E2 and optimal settings O1.

These correlations allow the machine learning model to predict and estimate optimal wireless device settings from explanatory variables (user wireless quality issues and device parameters). As long as the training dataset maintains the correlations, the model can make accurate and effective configuration suggestions.

[Explanation of the operation of the embodiment]
The overall flow of the function will be described using Figures 12 and 14. Figure 14 is a sequence diagram for explaining the operation of an embodiment of the setting support system of the present disclosure. First, the wireless device uses the natural language input function 102 to acquire wireless quality problems that the user is experiencing (step S1).

At the same time, the device parameter acquisition function 103 acquires the determined settings, status, etc. within the device (step S2).

Once the acquisition is complete, the acquired data is sent to the analysis server 201. The acquired data is then input into the natural language processing module 203 and device parameter encoding module 204, and the optimal settings are classified using the machine learning model 202 (step S3). The classification results are used by the setting presentation function 104, which presents a summary of the corresponding settings to the user. Note that because the classification results only indicate the optimal settings, the corresponding setting summary must be prepared in advance within the wireless device or the analysis server and presented using the setting presentation function 104.

After presenting the settings, the setting selection/completion input function 105 is used to ask the user whether they want to reflect the settings (step S4).

If the user selects to reflect the presented settings, they are prompted to enter any necessary information, such as a password, and the settings are then reflected using the settings reflection function 106 (step S5).

Finally, the wireless communication quality before and after the settings are reflected, and whether the proposed settings have been reflected, are sent as feedback to the analysis server 201 (step S6). Past communication quality is recorded in a past log, etc., so that the wireless communication quality after the settings are reflected can be obtained. In this way, appropriate settings are presented to the user based on the user's wireless quality issues and device parameters.

Next, continuous learning will be explained using Figure 15. Figure 15 is a flowchart for explaining the operation during continuous learning of an embodiment of the setting support system disclosed herein. Continuous learning is a mechanism for re-learning data newly obtained using some function, thereby maintaining or improving accuracy. In the present invention, when a user applies the presented optimal settings to a wireless device, a learning dataset including the results is used for continuous learning (step S1501).

Next, based on the value acquired by the wireless communication quality acquisition function 107, the improvement result of wireless communication quality is calculated as an improvement value between 0 and 1. This is calculated for each element representing wireless communication quality (delay, packet loss, etc.) (step S1502). Note that if the improvement value is negative, it is set to 0, and the calculation of the improvement value is determined according to each element representing wireless quality. For example, if the element representing delay, ping [ms], drops from 100 to 90, the improvement value for the latter will be higher than if it drops from 30 to 20; the improvement value is calculated according to the characteristics of the element.

Next, the average of the calculated improvement values (two if only delay and packet loss are considered) is calculated, and this score is used as the weight for the corresponding learning data set during continued learning (step S1503). This enables the machine learning model to focus on factors that have a significant impact on wireless communication quality during learning.

Finally, the user's wireless quality issues and device parameters are used as explanatory variables, the presented optimal settings as objective variables, and the calculated weights are used for continuous learning (step S1504). A specific example is presented. Consider a case where the user input is "wireless instability," and the output is two types, "band steering" and "change communication mode (such as turning IEEE 802.11ax settings ON)," with values of 0.55 and 0.45, respectively, and the band steering function is presented and reflected. After learning with a weight of 0.1, if the output is 0.6 and 0.4, then with a weight of 0.4, the output will be 0.7 and 0.3, which is more likely to present the "band steering function."

[Effect description]
The configuration support system according to the embodiment of the present disclosure can present appropriate settings to users without wireless knowledge by presenting settings based on machine learning models based on input about wireless quality issues. Furthermore, by using device parameters for machine learning predictions, it is possible to address individual user issues.

[Another Example 1]
In the present invention, the results obtained by the wireless device are used on the analysis server, but depending on the capacity of the entire machine learning model, it is also possible to complete the process using only the wireless device (Figure 16), or to select to perform only the learning in a separate environment (Figure 17).

When the process is completed using only the wireless device, the machine learning model and continuous learning are built into the wireless device without using an analysis server, as shown in Figure 16. In this case, it is necessary to reduce the performance of the natural language module, reduce the number of settings to be classified, and reduce the number of internal parameters for the entire machine learning model. Furthermore, since learning requires larger computational resources, it is necessary to either turn off other functions during non-use times and perform learning, or to prepare wireless devices with high computational processing capabilities in the first place. Furthermore, because there is a lack of training data, it is difficult to improve accuracy through continuous learning.

When classifying within the wireless device and performing learning in a separate environment, the machine learning model is incorporated into the wireless device as shown in Figure 17, and the continuous learning function is used on the analysis server. In this case, although not as important as above, attention must be paid to capacity and computing resources. Furthermore, updates to the machine learning model 202 after continuous learning can be incorporated into a firmware update.

[Another Example 2]
In the present invention, when a user inputs a wireless quality problem, a corresponding setting is presented. If the input is in natural language that has nothing to do with wireless quality, a mechanism that can respond appropriately will be described as another embodiment 2.

The training dataset is constructed so that the machine learning model can respond appropriately even when the user enters natural language input that is unrelated to wireless quality issues. Specifically, the training dataset is expanded to include natural language input that is unrelated to wireless quality (e.g., weather information or general conversation), along with the classification result of "no appropriate settings." By including such data in the training dataset, the machine learning model acquires the ability to respond appropriately to input that is unrelated to wireless quality issues and determine that "no appropriate settings" is present. As a result, the model does not make setting suggestions when the user enters unrelated content, preventing unrelated setting changes.

We will explain the correlation between "inappropriate settings" and "wireless quality problems." It can be said that a certain correlation exists between "inappropriate settings" and "wireless quality problems." However, this correlation does not indicate a relationship in which the two directly affect each other, but rather an inverse relationship. Specifically, the classification result of "inappropriate settings" means that it is applied to natural language input that is unrelated to wireless quality problems. In other words, when "inappropriate settings" is selected, it indicates that the presence of wireless quality problems is low or not related at all. Therefore, "inappropriate settings" and the presence of wireless quality problems have an inverse relationship with each other.

A part or all of the above disclosure may also be described as follows, but is not limited to the following.
[Appendix 1]
This is the same as the setting support device according to the first aspect described above.
[Appendix 2]
Preferably, the setting assistance device according to claim 1 further comprises a learning model generation unit that generates a learning model trained using learning data in which a setting problem and device parameters are used as input data and corresponding parameters corresponding to the setting problem and the device parameters are used as training data, and the calculation unit calculates the corresponding parameters using the learning model.
[Appendix 3]
Preferably, the setting support device according to claim 1 or 2, wherein the setting question acquisition unit acquires setting questions in natural language.
[Appendix 4]
The setting support device of Appendix 3, which is dependent on Appendix 2, preferably, wherein the setting problem acquisition unit vectorizes the setting problem written in natural language, and the learning model generation unit vectorizes the setting problem.
[Appendix 5]
Preferably, the setting support device of Appendix 2 further comprises a judgment unit that judges whether the corresponding parameters have been reflected in the device, and the learning model generation unit updates the learning model according to the judgment result of the judgment unit.
[Appendix 6]
Preferably, the learning model generation unit updates the learning model based on the corresponding parameters determined to be reflected in the device, the setting problem which is the input data for calculating the corresponding parameters, and the device parameters.
[Appendix 7]
This is the same as the setting support system according to the second aspect described above.
[Appendix 8]
The setting support system of Appendix 7, wherein the setting support server device further has a learning model generation unit that uses a setting problem and device parameters as input data and generates a learning model trained using learning data that uses corresponding parameters corresponding to the setting problem as teacher data, and the calculation unit calculates the corresponding parameters using the learning model.
[Appendix 9]
This is the same as the setting support method according to the third aspect described above.
[Supplementary Note 10]
This is the same as the program related to the fourth perspective mentioned above.
[Appendix 11]
The setting support device of Appendix 3, wherein the learning model calculation unit further uses a setting problem unrelated to the device parameters and corresponding parameters as input data and generates a learning model trained using learning data with an output indicating that there is no appropriate setting as training data.
Note that, like Supplementary Note 1, Supplementary Notes 9 and 10 can be expanded into Supplementary Notes 2 to 6.

The disclosures of the above-cited patent documents and other documents are incorporated herein by reference. Modifications and adjustments to the embodiments and examples are possible within the scope of the entire disclosure of the present invention (including the scope of the claims), and further based on the basic technical concepts thereof. Furthermore, various combinations and selections of the various disclosed elements (including each element of each claim, each element of each embodiment or example, each element of each drawing, etc.) are possible within the scope of the entire disclosure of the present invention. In other words, the present invention naturally includes various modifications and alterations that would be possible for a person skilled in the art in accordance with the entire disclosure and technical concepts, including the scope of the claims. In particular, with regard to the numerical ranges set forth herein, any numerical value or subrange included within that range should be construed as being specifically set forth, even if not otherwise specified.

1: User 2: Target device 3-5: Model, learning model 10: Setting support device 11: Setting problem acquisition unit 12: Device parameter acquisition unit 13: Calculation unit 14: Output unit 15: Learning model generation unit 16: Determination unit 17: Input data transmission unit 18: Corresponding parameter transmission unit 20: Setting device 30: Setting support server device 51: CPU
52: Memory 53: Input/output interface 54: NIC
55: Internal bus 101: Wireless device 102: Natural language input function 103: Device parameter acquisition function 104: Setting presentation function 105: Setting selection/complementary input function 106: Setting reflection function 107: Wireless communication quality acquisition function 201: Analysis server 202: Machine learning model 203: Natural language processing module 204: Device parameter encoding module 205: Integration module of both expressions 206: Classification module 207: Continuous learning function E1: Wireless quality problem E2: Device parameters O1: Wireless device settings

Claims (10)

a setting problem acquisition unit that acquires a setting problem, which is a problem related to the setting of a communication device, and which indicates an event that is inconvenient for a user ;
a device parameter acquisition unit that acquires device parameters including settings of the communication device;
a calculation unit that calculates corresponding parameters corresponding to the set problem based on the set problem and the device parameters using a learning model;
an output unit that outputs the corresponding parameters;
A setting support device having the above configuration. a learning model generation unit that generates a learning model trained using learning data in which the setting problem and the device parameters are used as input data, and corresponding parameters corresponding to the setting problem and the device parameters are used as training data;
the calculation unit calculates the corresponding parameters using the learning model;
The setting support device according to claim 1. The setting support device according to claim 1 or 2, wherein the setting question acquisition unit acquires the setting questions in natural language. the set question acquisition unit acquires the set question in natural language;
the set problem acquisition unit vectorizes the set problem written in the natural language;
The learning model generation unit is configured such that the set problem is a vectorized set problem.
The setting support device according to claim 2 . a determination unit that determines whether the corresponding parameter has been reflected in the communication device;
the learning model generation unit updates the learning model in accordance with the determination result of the determination unit.
The setting support device according to claim 2. the learning model generation unit updates the learning model based on the corresponding parameters determined to be reflected in the communication device, the set problem that is input data for calculating the corresponding parameters, and device parameters;
The setting support device according to claim 5. A setting support system including a setting device and a setting support server device,
The setting device includes:
a setting problem acquisition unit that acquires a setting problem, which is a problem related to the setting of a communication device, and which indicates an event that is inconvenient for a user ;
an apparatus parameter acquisition unit that acquires apparatus parameters including settings of the setting apparatus;
an input data transmitting unit that transmits the setting problem and the device parameters;
and
The setting support server device
a calculation unit that calculates corresponding parameters corresponding to the set problem based on the set problem and the device parameters using a learning model;
a corresponding parameter transmitting unit that transmits the corresponding parameters;
The setting device further comprises:
an output unit that outputs the corresponding parameters;
A setting support system having the above. The setting support server device
The method further includes a learning model generation unit that generates a learning model trained using learning data in which the set problem and device parameters are used as input data and corresponding parameters corresponding to the set problem are used as training data,
the calculation unit calculates the corresponding parameters using the learning model;
The setting support system according to claim 7. A setting support method for executing the following steps by a computer,
acquiring a setting problem, which is a problem related to the setting of a communication device, that indicates an event that is inconvenient for a user ;
obtaining device parameters including settings for the communication device;
calculating corresponding parameters corresponding to the set problem based on the set problem and the device parameters using a learning model;
outputting the corresponding parameters;
A setting support method having the above. A process of acquiring a setting problem, which is a problem related to the setting of a communication device, indicating an event that is inconvenient for a user ;
obtaining device parameters including settings for the communication device;
a process of calculating corresponding parameters corresponding to the set problem based on the set problem and the device parameters using a learning model;
a process of outputting the corresponding parameters;
A program that causes a computer to execute the following.