JP7401476B2 - Schedule creation support device and schedule creation support method - Google Patents

Description

The present invention relates to a schedule creation support device and a schedule creation support method.

The concept of so-called combinatorial optimization problems, which search for solutions that maximize or minimize desired parameters under given conditions, is useful for optimizing the placement and operation schedules of various resources such as workers and equipment, resolving traffic congestion, and global supply chains. It can also be applied to complex problems in the real world, such as reducing logistics costs.

On the other hand, since the number of solution candidates for such problems increases explosively, it is difficult to solve the problem within a practical amount of time unless a computer with appropriate computational power such as a supercomputer or a quantum computer is used.

For example, conventional technology related to quantum computers involves a computer that enables high-speed calculations for inverse problems and combinatorial optimization problems that require exhaustive search, and uses spin as a variable in the calculations to solve them. The problem is set up using spin-spin interactions and a local field that acts on each spin, and at time t = 0, all spins are directed in one direction by an external magnetic field, and at time t = τ, the external magnetic field becomes zero. The external magnetic field is gradually reduced, and each spin is allowed to evolve over time with its direction determined according to the effective magnetic field determined by the external magnetic field of each site at time t and all the interactions between spins. Techniques have been proposed in which the system maintains almost its ground state by not completely aligning with the effective magnetic field but in a quantum mechanically corrected direction (see Patent Document 1).

In addition, as a conventional technology for generating work schedules for personnel, a communicator is used for the purpose of creating a schedule for a communicator whose skills match the anticipated inquiry content based on the communicator's skills from among communicators who can be scheduled. A method for creating a data schedule (see Patent Document 2) has been proposed.

This technique is a method for a computer to create a schedule for a plurality of communicators in a contact center, the computer storing at least data indicating the skills of each of the plurality of communicators in a storage section, and the computer a step of receiving, from a contact center administrator, a designation input of the number of communicators to be assigned for each of the plurality of skills of the communicators for each business time period; and a step of receiving data indicating the number of people who have received the designated input. The method further includes the steps of further storing data in the storage unit for each business time period, and creating a schedule for the plurality of communicators based on the data indicating the stored number of people for each business time period. It is something that

In addition, the shift information that stores the shift information in which the first person working during the target period is assigned to the target period is aimed at facilitating sudden changes in work shifts in facilities such as childcare facilities. A shift management device (see Patent Document 3) has also been proposed, which includes a storage unit and a shift creation unit that allocates a predetermined number of spare second personnel to the first personnel for the target period of the shift information. There is.

WO2016/157333 Japanese Patent Application Publication No. 2008-186203 JP2015-212881A

However, no form has been proposed in which the quantum computer technology described above is appropriately applied to the task of creating an overall schedule when many people work together.

For example, in schedule creation work at a call center, it is necessary to create weekly and monthly shift schedules for several hundred operators. Currently, experienced personnel create schedules manually based on predetermined rules (for example, shift patterns such as early shifts and late shifts are arranged in a fixed order and frequency for the relevant period). Ta.

However, prior to creating such a schedule, it is possible to estimate the amount of work and the number of required personnel for each time period by employing an appropriate algorithm such as machine learning.

If we were to estimate the number of personnel required for each time period based on a model obtained through machine learning, the value would not be an integer value determined by humans based on experience, but a decimal number determined based on the amount of work. There will also be cases where this happens.

Suppose that in this way, the number of personnel required for a certain task in a certain time period can be estimated to be "3.4 people." In such cases, the decision made by the person in charge of creating the schedule and the on-site supervisor has been to allocate four staff members with appropriate skills to deal with the situation. In other words, redundant staffing was frequently carried out, exceeding the needs that were estimated with high accuracy using machine learning and were considered to be more closely related to actual needs.

Furthermore, although each staff member working under such an operation often has multiple skills, no technique has been proposed to create a schedule that allows the staff member to work efficiently according to the skills they possess. Even if on-site supervisors sometimes assigned staff to multiple tasks at their discretion, it was always based on intuition and experience.

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide a technique for efficiently creating a schedule in which the necessary personnel are allocated for each task in each time period without waste.

The schedule creation support device of the present invention that solves the above problems has the following advantages: the total working length of each person working together in a predetermined task in a predetermined period, the required number of people who can include decimals at each timing in the period, and A storage unit that stores information on constraints when allocating personnel, the total working length in the period, the required number of people, and a constraint function that is minimized when the constraints are satisfied, as terms. and a calculation unit that calculates an Ising model in which the operating rate of each of the personnel is set as a spin, and the sensitivity between variables in the constraint condition function is set as the strength of interaction between the spins, with respect to the objective function included. , the calculation unit outputs to a predetermined device a schedule that defines the operating rate of each of the personnel at each timing of the predetermined period based on the result of the calculation.

Further, the schedule creation support method of the present invention includes: the total working length of each person working together in a predetermined period in a predetermined period; the required number of people who may include a decimal number at each timing in the period; and the assignment of each person to the task. an information processing device that includes a storage unit storing information on the total working length in the period, the required number of people, and a constraint function that is minimized when the constraint conditions are satisfied; Regarding the objective function included as a term, an Ising model is calculated in which the utilization rate of each personnel is set as spin, and the sensitivity between variables in the constraint condition function is set as the strength of interaction between the spins, and the Based on the results, a schedule defining the operating rate of each of the personnel at each of the timings in the predetermined period is output to a predetermined device.

According to the present invention, it is possible to efficiently create a schedule that allocates the necessary personnel for each task in each time slot without waste.

FIG. 1 is a network configuration diagram including a schedule creation support device according to the present embodiment. It is a diagram showing an example of the hardware configuration of the schedule creation support device in this embodiment. It is a figure showing an example of a timing chart of this embodiment. It is a figure showing an example of a flow in this embodiment. FIG. 3 is a diagram showing an example of the data structure of a basic information table according to the present embodiment. FIG. 3 is a diagram illustrating an example data structure of a constraint table according to the present embodiment. FIG. 3 is a diagram illustrating an example data structure of a constraint table according to the present embodiment. It is a figure showing an example of a flow of a schedule creation support method in this embodiment. It is a figure which shows the example of a screen in this embodiment. It is a figure which shows the example of a screen in this embodiment. It is a figure which shows the example of a screen in this embodiment. It is a figure which shows the example of a screen in this embodiment. It is a figure which shows the example of a screen in this embodiment. 1 is a diagram showing an example of a CTI system in this embodiment. It is a figure showing an example of the management table in this embodiment. It is a figure showing an example of a flow in this embodiment.

<About the annealing machine>
As shown in the above-mentioned Patent Document 1, the present applicant has developed quantum computing technology and has attempted to solve various problems in, for example, exhaustive search problems (including the concept of combinatorial optimization problems) based on big data. .

In general, there are high expectations for quantum computers to solve such exhaustive search problems. Quantum computers are made up of basic elements called quantum bits, which can simultaneously realize "0" and "1". Therefore, it is possible to simultaneously calculate all solution candidates as initial values, and there is a possibility of realizing an exhaustive search. However, quantum computers need to maintain quantum coherence throughout the entire computation time.

Among these, a method called adiabatic quantum computation has started to attract attention (Reference: E. Farhi, et al., "A quantum adiabatic
evolution al gorithm applied to random
instances of an NP-complete problem, "S
science292, 472 (2001). ).

This method transforms the problem so that the ground state of a certain physical system becomes the solution, and attempts to obtain the solution by finding the ground state.

Let H^ _p be the Hamiltonian of the physical system in which the problem is set. However, at the start of the calculation, instead of using H^ _p as the Hamiltonian, we use another Hamiltonian H^ ₀ whose ground state is clear and easy to prepare. Next, take enough time to shift the Hamiltonian from H^ ₀ to H^ _p . If enough time is spent, the system will remain in the ground state, and the ground state of the Hamiltonian H^ _p will be obtained. This is the principle of adiabatic quantum computation. If the calculation time is τ, the Hamiltonian becomes the formula (1),

[Formula 1]

A solution is obtained through time evolution based on the Schrödinger equation of equation (2).
[Formula 2]

Adiabatic quantum computation can also be applied to problems that require an exhaustive search, and the solution is reached in a unidirectional process. However, if the calculation process needs to follow the Schrödinger equation of equation (2), it is necessary to maintain quantum coherence as in a quantum computer.

However, whereas a quantum computer repeats a gate operation between one or two qubits, adiabatic quantum computation involves making the entire qubit system interact all at once, and the concept of coherence is different.

For example, consider the operation of a gate on a certain quantum bit. At this time, if there is an interaction between that qubit and other qubits, it will cause decoherence, but in adiabatic quantum computation all qubits are made to interact at the same time, so in a case like this example does not become decoherent. Reflecting this difference, adiabatic quantum computation is thought to be more robust against decoherence than quantum computers.

As mentioned above, adiabatic quantum computation is effective for difficult problems that require exhaustive search. Then, the spin is used as a variable in the calculation, and the problem to be solved is set using spin-spin interactions and local fields that act on each spin.

At time t=0, all spins are directed in one direction by an external magnetic field, and the external magnetic field is gradually reduced so that it becomes zero at time t=τ.

Each spin evolves over time with its direction determined in accordance with the effective magnetic field determined by the external magnetic field of each site at time t and all the effects of interactions between spins.

At this time, the spin direction is not completely aligned with the effective magnetic field, but is quantum mechanically corrected so that the system maintains almost its ground state.

Additionally, a term (relaxation term) that maintains each spin in its original direction during time evolution is added to the effective magnetic field to improve the convergence of the solution.

The schedule creation support device in this embodiment is assumed to be an annealing machine that performs the above-mentioned adiabatic quantum calculation, but it is of course not limited to this. Any method that can be solved is applicable.

Specifically, the annealing method includes not only hardware implemented using electronic circuits (digital circuits, etc.), but also methods implemented using superconducting circuits. Furthermore, hardware that implements the Ising model using a method other than the annealing method may be used. Examples include laser network methods (optical parametric oscillation) and quantum neural networks. Furthermore, although some of the ideas are different as described above, the present invention can also be realized in a quantum gate method in which calculations performed using the Ising model are replaced with gates such as Hadamard gates, rotation gates, and control NOT gates.

<Network configuration>
Embodiments of the present invention will be described in detail below with reference to the drawings. FIG. 1 is a network configuration diagram including the schedule creation support device 100 of this embodiment.
The schedule creation support device 100 shown in FIG. 1 is a computer device that efficiently creates a schedule that allocates the necessary personnel for each task in each time period without waste. Specifically, an annealing machine is assumed as an example. do.

However, the outline of the annealing machine is as already described based on Patent Document 1, and details such as its specific configuration and operation will be omitted as appropriate (the same applies hereinafter).

The schedule creation support device 100 of this embodiment is connected to a user terminal 200 for data communication via an appropriate network 10 such as the Internet.

The user terminal 200 described above is a terminal that receives a schedule proposal from the schedule creation support device 100.

Specifically, the user of this user terminal 200 can be assumed to be an organization such as a financial institution, an insurance company, or a major manufacturer, which is a business that has many operators and performs call center operations. Alternatively, it can also be assumed that a medical institution or a care provider has many nurses and care staff to carry out nursing and care work for patients and the like.

In any case, if the business intends to allocate the necessary number of personnel of a suitable size and each with one or more skills to each date and time period, and to carry out the work without excess or deficiency as a whole, This may apply to the above users. In other words, it can be said that the present invention is applicable to the operations of such businesses.

As a concrete example, when creating schedules at a bank's call center, the reality is that a person with a certain level of experience manually creates monthly shift schedules for all hundreds of operators. . In other words, the work tends to be individualized and is a manual task.

However, prior to creating such a schedule, it is possible to estimate the amount of work and the number of required personnel for each time period by employing an appropriate algorithm such as machine learning.

For example, based on the amount of work that has occurred in the past in each time period, we can identify the correlation between the amount of work and various factors such as date and time, day of the week, weather, etc., and based on this, we can determine the amount of work that will occur in each time period in the future. and estimate its contents. The number of personnel required to handle the workload can be calculated, for example, by dividing the workload of the job requiring the skill by the workload of each staff member with the skill. As a result, there are cases where the number of required personnel is not an integer but a decimal number.

If it is estimated that the number of staff required for a certain task during a certain time period is ``3.4 people'', the person in charge of creating the schedule and the on-site supervisor may decide to deploy ``4'' staff with the necessary response skills. The conventional method of operation has been to deal with this situation. In other words, redundant personnel assignments that exceeded actual needs were frequently carried out.

Furthermore, although each staff member working under such an operation often has multiple skills, no technique has been proposed to create a schedule that allows the staff member to work efficiently according to the skills they possess. Even if on-site supervisors sometimes assigned staff to multiple tasks at their discretion, it was always based on intuition and experience.

The above-mentioned obstacles are due to mathematical difficulties and the absence (strictly speaking, immaturity) of technology to solve the mathematical difficulties. For example, spreadsheet software can mathematically solve linear constraints quickly, but it takes an enormous amount of time to solve nonlinear constraints.

Conventionally, when creating a schedule, the amount of calculations increases rapidly due to the increase in nonlinear constraints related to elements, such as each personnel, and it takes a long time to complete the calculations, but using an annealing machine By employing the schedule creation support device 100, it becomes possible to perform calculations without depending much on the increase in the number of elements.

<Hardware configuration>
Further, the hardware configuration of the schedule creation support device 100 of this embodiment is shown in FIG. 2 as follows.

That is, the schedule creation support device 100 includes a storage section 101, a memory 103, a calculation section 104, and a communication section 105.

Among these, the storage unit 101 is configured with an appropriate nonvolatile storage element such as an SSD (Solid State Drive) or a hard disk drive.

Furthermore, the memory 103 is composed of a volatile storage element such as a RAM.

Further, the calculation unit 104 is a CPU that reads out and executes the program 102 held in the storage unit 101 into the memory 103, performs overall control of the apparatus itself, and performs various determinations, calculations, and control processes.

Further, the communication unit 105 is assumed to be a network interface card or the like that connects to the network 10 and handles communication processing with the user terminal 200.

Note that when the schedule creation support device 100 is a stand-alone machine, it is preferable to further include an input device that accepts key inputs and voice inputs from the user, and an output device such as a display that displays processed data.

Furthermore, in the storage unit 101, at least a basic information table 125 and a constraint table 126 are stored, in addition to the program 102 for implementing the functions necessary for the schedule creation support device of this embodiment. However, details regarding these tables will be described later.

Further, the program 102, that is, the algorithm that implements the operation as an annealing machine, holds information on the Ising model 1021, which is the problem to be solved. This Ising model 1021 is set in advance by an administrator or the like based on the tasks and personnel to whom information is provided, and other various information that affects them.

Note that the adiabatic quantum computation described in the overview of annealing machines is also called quantum annealing, and is an extension of the classical annealing concept to quantum mechanics. In other words, adiabatic quantum computation is originally capable of classical operation, and can be interpreted as adding quantum mechanical effects to improve performance in terms of speed and accuracy of solutions. Therefore, in the present invention, the calculation section itself is classical, and by introducing parameters determined by quantum mechanics into the calculation process, a calculation method and device that is classical but includes quantum mechanical effects is realized. However, it is of course possible to adopt a configuration in which the calculation section is configured with a quantum computer.

Based on the above concept, in the following example, we will describe a classical algorithm for obtaining a ground state as a solution and a device for realizing it, while explaining its relationship with adiabatic quantum computation.

The schedule creation support device 100 based on this assumption assumes that N variables sj ^z (j=1, 2,..., N) take a range of -1≦sj ^z ≦1, and that the local field gj and the interaction between variables Jij The assignment is set by (i, j = 1, 2, ..., N).

In addition, the calculation unit 104 divides the time into m and calculates discretely from t=t ₀ (t ₀ =0) to tm (tm = τ), and calculates the variable sj ^Z (tk) at each time tk. Then, using the value of the variable Sj ^z (tk-1) (i = 1, 2, . . . , N) at the previous time tk-1 and the coefficient 9pina or 9pinb of the relaxation term, Bj ^z (tk) = {ΣiJijsi ^z (tk-1)+gj+sgn(sj ^z (tk-1))・9pina}・tk/τ or Bj ^z (tk)={ΣiJiJsj ^z (tk-1)+gj+9pinb . sj ^Z (tk-1)}・tk/τ, determine the function f so that the range of the above variable sj ^z (tk) is -1≦sj ^z (tk)≦1, and calculate sj ^z (tk) = f(Bj ^z (tk), tk), and as the time step advances from t=t ₀ to t=tm, the above-mentioned variable sj ^z approaches -1 or 1, and finally if sj ^z <0, If sj ^zd =-1 and Sj ^z >0, a solution is determined by setting sj ^zd =1. However, if it is appropriate for the final solution sj ^zd to be a real number, the solution may be determined by setting sj ^zd as a real number whose range is [-1, 1].

The coefficient gpinb is, for example, a value between 50% and 200% of the average value of |Jij|. Furthermore, regarding the local field gj of the task setting, it is also possible to add a correction term δgj' to gj' only for a certain site j' and increase the size of gj' only for that site j'. Further, the correction term δgj' is, for example, a value from 10% to 100% of the average value of |Jij|.

Next, we will explain the basic principles of annealing machines by starting from a quantum mechanical description and moving to a classical form.

The ground state search problem for the Ising spin Hamiltonian given by Equation (3) includes a class of problems called NP-hard, and is known to be a useful problem (Reference: F. Barahona, "On the computational comp lexity of Isingspin glass models,” J. Phys.
A: Math. Gen. 15, 3241 (1982). ).

[Formula 3]

Jij and gj are task setting parameters, and σ^ ^Z is the z component of the Pauli spin matrix and takes an eigenvalue of ±1. i and j represent spin sites. Ising spin is a variable that can take only ±1 as a value, and in equation (3), the eigenvalue of σ^ ^z is ±1, making it an Ising spin system.

The Ising spin in Equation (3) does not need to be a literal spin, and may be any physical spin as long as the Hamiltonian is described by Equation (3).

For example, it is possible to associate each operating rate and each proficiency level for each skill of each person with spin ±1, it is also possible to associate high and low of a logic circuit with ±1, and it is also possible to associate the vertical and horizontal polarization of light. It is also possible to associate waves with ±1 and to associate phases of 0 and π with ±1.

In the method illustrated here, the calculation system is prepared in the ground state of the Hamiltonian given by equation (4) at time t=0, similar to the adiabatic quantum calculation.

[Formula 4]

γ is a proportionality constant determined by the magnitude of the external field uniformly applied to all sites j, and σ^j ^x is the x component of the Pauli spin matrix. If the calculation system is spin itself, the external field means the magnetic field.

Equation (4) corresponds to applying a transverse magnetic field, and the ground state is when all spins are oriented in the x direction (γ>0). The Hamiltonian in the problem setting was defined as an Ising spin system with only the z component, but the x component of spin appears in equation (4). Therefore, the spin in the calculation process is vector-like (Bloch vector) rather than Ising. At t=0, it starts with the Hamiltonian of Equation (4), but as time t progresses, the Hamiltonian is gradually changed, and finally the Hamiltonian described by Equation (3) is obtained, and its ground state is obtained as a solution.

[Formula 5]

Here, σ^ represents the three components of the Pauli spin matrix as a vector. The ground state is when the spin points in the direction of the magnetic field, and can be written as <σ^>=B/|B|, where <·> is the quantum mechanical expectation value. In an adiabatic process, the ground state is always maintained, so the direction of the spin always follows the direction of the magnetic field.

The above discussion can be extended to multi-spin systems. At t=0, the Hamiltonian is given by equation (4). This means that the magnetic field Bj ^x =γ is uniformly applied to all spins. At t>0, the x component of the magnetic field gradually weakens, and Bj ^X =γ(1−t/τ). Regarding the z component, since there is an interaction between spins, the effective magnetic field is expressed as equation (6).

[Formula 6]

Since the spin direction can be defined by <σ^ ^z >/<σ^ ^X >, if the spin direction follows the effective magnetic field, the spin direction is determined by equation (7).

[Formula 7]

Although equation (7) is a quantum mechanical description, it takes an expected value, so unlike equations (1) to (6), it is a relational equation regarding classical quantities.

In the classical system, there is no non-local correlation (quantum stitching) of quantum mechanics, so the direction of the spin should be completely determined by the local field at each site, and Equation (7) determines the behavior of the classical spin system. In quantum systems, equation (7) has to be modified because there is non-local correlation, but this will be discussed later.Here, in order to describe the basic form of the invention, we will discuss the classical system determined by equation (7). Describe.

FIG. 3 shows a timing chart (1) for obtaining the ground state of the spin system. Since the description in FIG. 3 concerns classical quantities, the spin at site j is expressed by sj instead of σ^ _j . Further, accordingly, the effective magnetic field Bj in FIG. 3 is a classical quantity. At t=0, a rightward effective magnetic field Bj is applied to all sites, and all spins sj are initialized to the right.

As time t passes, the magnetic field in the z-axis direction and the interaction between the spins are gradually applied, and finally the spins become in the +z direction or -z direction, and the z component of the spin Sj becomes sj ^z = +1 or -1 becomes. Ideally, the time t is continuous, but in an actual calculation process, it can be made discrete to improve convenience. In the following, we will discuss the discrete case.

The spin illustrated here is a vector-like spin because not only the z component but also the x component is added. The behavior as a vector can also be understood from FIG. The y component has not appeared so far, but this is because the y component of the external field does not exist because the direction of the external field is set to the xz plane, and therefore <σ^ ^Y >=0.

The spin of the calculation system is assumed to be a three-dimensional vector of magnitude 1 (this is called a Bloch vector, and the state can be described by a point on the spherical surface), but in the example shown in the figure, the axis is taken as a two-dimensional vector. (states can be described by points on a circle).

Furthermore, since γ is constant, Bj ^x (t)>0 (γ>0) or Bj ^x (t)<0 (γ<0) holds true. In this case, the two-dimensional spin vector can be described only by a semicircle, and if sj ^z is specified by [-1, 1], the two-dimensional spin vector is determined by one variable, sj ^z . Therefore, in this example, spin is a two-dimensional vector, but it can also be expressed as a one-dimensional continuous variable with a range of [-1, 1].

In the timing chart of FIG. 3, the effective magnetic field is determined for each site at time t=tk, and using that value, the spin direction at t=tk is determined by equation (8).

[Formula 8]

Equation (8) is a rewrite of Equation (7) into the notation related to classical quantities, so it does not have the <・> symbol. Next, the effective magnetic field at t=tk+l is determined using the spin value at t=tk. If the effective magnetic field at each time is specifically written, it becomes Equations (9) and (10).

[Formula 9]

[Formula 10]

Hereinafter, the spin and the effective magnetic field are determined alternately according to the procedure schematically shown in the timing chart of FIG.

In the classical system, the magnitude of the spin vector is 1. In this case, each component of the spin vector is written as sj ^z (tk ⁾ = sin θ, sj ^x (tk) = COS θ using the parameter θ defined as tan θ = Bj z (tk)/Bj ^x (tk). Ru.

Rewriting this, sj ^z (tk) = sin(arctan(Bj ^z (tk)/Bj ^x (tk))), sj ^x (tk) = cos(arctan(Bj ^z (tk)/Bj ^x (tk) )).

As is clear from equation (9), the only variable in Bj ^x (tk) is tk, and τ and γ are constants. Therefore, Sj ^z (tk)=sin(arctan(Bj ^z (tk)/Bj ^x (tk))) and sjx(tk)=cos(arctan(Bj ^z (tk)/Bj ^x (tk))) are Bj Generalizations such as sj ^{z (} tk) = f1 (Bj ^z (tk), tk) and sj ^x (tk) = f2 (Bj ^z (tk), tk) as functions with z (tk) and tk as variables You can also express

Since spin is described as a two-dimensional vector, two components, sj ^z (tk) and sj ^x (tk), appear, but if Bj ^z (tk) is determined based on equation (10), sj ^x (tK) is not necessary.

This corresponds to the fact that the spin state can be described only by sj ^z (tk) whose range is [-1, 1]. The final solution sj ^zd is generally extracted by sj ^zd = -1or1, if sj ^z (τ) > 0, sj ^zd = 1, and if sj ^z (τ) < 0, sj ^zd = -1. shall be. However, if it is appropriate for the final solution sj ^zd to be a real number, sj ^zd may be extracted as a real number whose range is [-1, 1].

FIG. 4 shows a flowchart summarizing the above algorithm. Here, tm=τ. Each step s1 to s9 in the flowchart in FIG. 4 corresponds to processing at a certain time in the timing chart in FIG. 3 from time t=0 to t=τ. That is, steps s2, s4, and s6 of the flowchart correspond to the above equations (9) and (10) at t=t1, tk+l, and tm, respectively. The final solution is determined in step s8 by setting sj ^zd =-1 if sj ^z <0, and setting sj ^zd =1 if sj ^z >0 (s9). However, if it is appropriate for the final solution to be a real number, sj ^zd may be determined as a real number whose range is [-1, 1]. Note that the flow in FIG. 4 shows a general example, and the description of extracting as a real number is omitted.

So far, we have shown how the problem is solved when it is expressed by equation (3). Next, a specific example will be given to explain how a specific problem is expressed by equation (3) including a local field gj and an interaction between variables Jij (i, j=1, 2, . . . , N).

The specific problem here, that is, the Ising model 1021, includes, for example, the total number of working days in a given month for each operator involved in call center work, the required number of people who may include decimals for each day in the month, and the assignment of each operator to call center work. Based on the information on the constraint conditions, the total number of working days in the month, the above-mentioned required number of people for each day in the month, and the minimum constraint function when the above-mentioned constraint conditions are satisfied. Regarding the objective function included as , an Ising model is assumed in which the operating rate of each operator is set as the spin, and the sensitivity between variables in the constraint function is set as the strength of the interaction between the spins.

In this case, the local field gj is a variable in the above-mentioned total number of working days in the month, the required number of people who can include decimals for each day in the month, and the constraint function that is minimized when the above-mentioned constraints are satisfied. It is assumed that the value of is set as the degree of influence on the objective function.

Other examples of the Ising model include, for example, the total number of working days in a given month for each operator involved in call center work, the required number of people who can include a decimal number for each required skill on each day of the month, and the number of workers involved in call center work. Based on the information on the constraints when allocating operators, the total number of working days in the month, the required number of people for each day in the month, and the minimum constraint function when the above constraints are satisfied. Regarding the objective function that includes , as a term, it is also possible to assume an Ising model in which the utilization rate for each skill of each operator is set as the spin, and the sensitivity between variables in the constraint function is set as the strength of the interaction between the spins. can.

In this case, the local field gj is the total number of working days in the month, the required number of people for each day of the month, which may include a decimal number, and the minimum constraint when the above constraints are satisfied. It is assumed that the value of the variable in the objective function is set as the degree of influence on the objective function.

In addition, other examples of the Ising model include, for example, constraints regarding rare tasks that require important skills, such as the number of personnel required to handle the rare tasks, and the time slot for handling the rare tasks. Regarding the objective function that includes the constraint function as a term, assume an Ising model in which the utilization rate for each skill of each operator is set as the spin, and the sensitivity between variables in the constraint function is set as the strength of the interaction between the spins. You can also do it.

In this case, the local field gj is the total number of working days in the month mentioned above, the required number of people who can include a decimal number for each required skill for each day in the month, the required number of personnel to handle the rare work, and the number of people who can handle the rare work. It is assumed that the value of the variable in the time period and the constraint function that becomes minimum when the above-mentioned constraint is satisfied is set as the degree of influence on the objective function.

In addition, as another example of the Ising model, for example, the utilization rate and proficiency level for each skill of each person is taken as spin, and the sensitivity between variables in the above-mentioned constraint function is taken as the strength of interaction between spins. It is also possible to assume a set Ising model.

Through the above considerations, we specifically set the inter-variable interaction Jij (related to each term of the objective function described above) and the local field gj, and search for the ground state of the Ising model 1021 expressed by equation (3). , that is, the objective function consisting of the total number of working days in the month mentioned above, the required number of people who can include decimals for each day in the month, and the constraint function that is the minimum when the above constraint conditions are satisfied is the minimum. Through the search for the base state, each operator's work shift, that is, the entire schedule for the month in call center work is identified.

Note that the Ising model and annealing method only calculate ``minimizing the objective function.'' Therefore, if there are constraints that need to be satisfied when minimizing the objective function, it is necessary to add them to the objective function in some way.

for example,
[Formula 11]
Let's consider the following constraint. If we convert this constraint condition into "a function that is minimized when the constraint condition is satisfied", the following formula will be obtained.

[Formula 12]

Since the squared part is always a positive value, this formula has a minimum value only when the content of the square is 0. The content becomes 0 only when ΣXi - A = 0, so if you solve the optimization problem where this function is minimized, you will automatically obtain a solution that satisfies ΣXi = A. Become.

Further, for example, in the above-described annealing method, it is necessary to include items to be used as constraints in the objective function, so the objective function and the constraints are treated with the same importance.

For example, suppose we have the following optimization problem.
[Formula 13]

Changing this to the formulation for annealing results in the following.
[Formula 14]

Here, P and Q are constants, and are factors that determine which term is minimized preferentially. For example, if you want to minimize all three terms equally (that is, solve the problem without biasing the strength of the constraints), you can change the values to balance the terms, such as making P and Q equal. Make settings.

On the other hand, if the problem is set to ``strictly follow the second term's constraint, but do not place much emphasis on the third term's constraint,'' set the value of P, which is the coefficient of the term to be emphasized, to be larger than the value of Q. By doing so, you will get the desired solution.

In this way, according to the annealing method, it is possible to prioritize constraints and set constraints that are not given much importance to be "satisfied as much as possible."

Note that various settings related to the Ising model are appropriately performed in accordance with existing general concepts in accordance with each condition and information for schedule generation in this embodiment.

<Specific example of spin (variable) settings>
A variable x_(i,j,k, m) (see Figure 5). Basically, each of these variables x_(i,j,k,m) is assigned to each spin in the Ising model. Furthermore, when considering the proficiency level n of skills corresponding to each work type,
Variable x_(i ,j,k,m,n). Basically, each of these variables x_(i,j,k,m,n) is allocated to each spin in the CMOS annealing machine. In addition, if each person has a single level of proficiency for each skill, the sensitivity between variables in the constraint function is calculated using variables x_(i,j,k,m) that do not take proficiency into account It can also be set by multiplying the proficiency level.

If you do not consider "work type", this variable is defined as "a certain employee i has an utilization rate of m% during a certain time period j".
It is a variable that gives whether or not to do business. In other words, in the cells containing 1 in Table 1 below, operations will be carried out at an operating rate of m%, and in the cells containing 0, operations will not be carried out.

On the other hand, when considering "type of work" and "operating rate," in the example of a call center, "inbound work" corresponds to k=1, and "outbound work" corresponds to k=2. Then, there are two variables (x_1,1,1,m,n and x_1,1,2,m,n) corresponding to the top left cell in Table 1 above. Since these two variables each take a value of 0 or 1, there are four possible combinations as shown in Table 2 below.

When considering "work type" and "utilization rate", in Table 2 above, work type "inbound" corresponds to k=1, work type "outbound" corresponds to k=2, and the combination of two variables represents one cell on a table (hereinafter referred to as "timetable") that represents the shifts of each person per unit time. In other words, if the two variables (x_(i,j,1,m,n) and x_(i,j,2,m,n)) belonging to a certain cell are both 0, it is a "rest"; If ,j,1,m,n) is 1 and all x_(i,j,2,m,n) are 0, "inbound" is executed with an availability rate of m%×n, and all x_(i,j, 1, m, n) is 0 and one x_(i, j, 2, m, n) is 1, it means that "outbound" is executed at an operating rate of m% x n. Here, if x_(i,j,1,m,n) and x_(i,j,2,m,n) are both 1, then ``the same person can perform two different tasks in the same time period, respectively. This will be implemented with an operating rate of m% x n.

<Data structure example>
Next, various information used by the schedule creation support device 100 of this embodiment will be explained. FIG. 5 shows an example of the basic information table 125 in this embodiment.

The basic information table 125 of this embodiment includes the total number of working hours in a predetermined month of each operator who collaborates in call center operations, and the number of hours worked (by those who have the corresponding skills) for each task in each time period of each day in the month. This table stores information about the number of people required.

Among these, the value of the required number of people can be estimated using an appropriate algorithm such as machine learning. In this case, the schedule creation support device 100 stores, for example, information on the amount of past work that occurred in each time period in the storage unit 101, and based on this information, schedules for various things such as date and time, day of the week, weather, etc. A model is generated by identifying the correlation between factors and the amount of each task generated using a machine learning algorithm. For example, if a certain time period is given as input to a model that has learned this correlation, it is possible to estimate the amount of each task that is expected to occur during that time period.

In addition, if the schedule creation support device 100 holds information on the amount of each task that actually occurred in the past and the number of personnel who handled the task without any problems, the schedule creation support device 100 similarly performs machine learning on the correlation between them. It is also possible to envisage a configuration in which a model is generated, and information on the amount of each task estimated as described above is added to this model to estimate the number of required personnel.

Alternatively, the required number of staff may be calculated by dividing the amount of each work estimated as described above by the average amount of work handled by each staff member with a certain skill.

It should be noted that other methods of calculating the required number of people can be envisaged. For example, it is also possible to use a method of calculating a decimal number of required personnel by drawing an approximation curve from the required number of people (an integer) calculated using the Erlang formula for estimating the number of required operators at a call center.

In the example of this embodiment, the time interval is set to 1 hour, and the optimal combination satisfies the required number of people for each job specified in the horizontal row in each time period, and protects the number of working hours (number of working periods) for each operator. This becomes a problem. The problem is that the finer the time interval, the more cells required for calculation (cells = factors in this table), and the more calculation power is required.

Furthermore, since there are a large number of cells, when setting constraints such as increments of the operating rate and proficiency level of each operator, and various prohibited patterns, the number of constraint function functions corresponding to the constraint conditions becomes enormous. However, if an annealing machine is employed, a favorable result or schedule can be specified within a realistic amount of time.

Note that the example illustrated in the basic information table 125 in FIG. 5 is limited for convenience of explanation, and it is assumed that information on various other events is stored (the same applies below).

Further, FIGS. 6 and 7 show an example of the constraint table 126 in this embodiment. The constraint table 126 of this embodiment is a table that stores information on each constraint condition when assigning each operator to the above-mentioned call center work.

The data structure uses the identification information of each constraint (“#” in the figure) as a key, and includes the contents of the constraint, an implementation example in an optimization solver, and an implementation example in an annealing machine (constraint function). . It is a collection of records consisting of such data.

In this embodiment, in addition to an implementation example in an annealing machine, that is, a constraint function, an example of a function when implementing the constraint in an optimization solver is also displayed for comparison with the conventional technology.

<Flow example>
Hereinafter, the actual procedure of the schedule creation support method in this embodiment will be explained based on the drawings. Various operations corresponding to the schedule creation support method described below are realized by a program that the schedule creation support device 100 reads into a memory or the like and executes. This program is composed of codes for performing various operations described below.

FIG. 8 is a diagram showing a flow example of the schedule creation support method in this embodiment. In this case, the schedule creation support device 100 calculates, as the Ising model 1021 to be processed, the total number of work shifts of each operator in a certain month, the number of people required for each job on each day of the month, and the constraint function. Regarding the objective function included as a term, an Ising model 1021 is calculated in which the operating rate and proficiency level for each skill of each operator are set as spins, and the sensitivity between variables in the constraint condition function is set as the strength of interaction between spins. shall be.

Here, the utilization rate is the ratio of working time for a certain task (that is, a task performed with a specific skill) within a certain unit time. For example, an operating rate of 75% when the unit time is 1 hour indicates a situation in which a worker works on a certain task for 75% of the hour, that is, 45 minutes, and takes a 15-minute break. In this case, combinations of skills, that is, task ratios (hereinafter referred to as "labels") are prepared in advance, and the optimal one is selected from among them.

For example, as shown in Table 3 below, we prepare five types of skill combinations (= "labels") and use an optimization method to determine which label should be assigned to each person and each time period. select. In addition, in the example of Table 3, the number of skills is 2 and the utilization rate meter is 100%, but the number is not limited to this.

In addition, proficiency is an index in which, for example, if a person with average work efficiency is ``1.0'', the proficiency level of a person who performs work 20% more efficiently is ``1.2''. be. Or, if the business is such that ``two newcomers do the work of one veteran,'' then the proficiency level of the new employee should be set to ``0.5'' and the proficiency level of the other veterans should be set to ``1.'' 0”.

In this way, in the method of calculating the allocated number of people (person-days/person-hours) using the operating ratio and proficiency of each operator's job, that is, each skill, the definitions of variables and constants are, for example, as follows. x_(i,j, k), the utilization rate of skill s at label k is r_(k,s), and the proficiency level (for example, information in Table 4) of person i when implementing skill s is e_(i,s). .

In addition, to calculate the number of people to be assigned, for example, if the number of people (person-days/person-hours) assigned to skill s on day j (time period j in the case of a timetable) is W_(j,s), then W_(j,s ), and the conditional expression 16 that indicates "assign the required number of people (person-days/person-hours) N _js or more for skill s on day j (time zone j in the case of a timetable)" do.

[Formula 15]

[Formula 16]

Therefore, the schedule creation support device 100 acquires the desired conditions of each operator from the user terminal 200 and stores them as constraints in the constraint table 126 (s10). Alternatively, at this time, designation of a constraint that matches the desired constraint from the constraint table 126 may be accepted from each operator.

Furthermore, the schedule creation support device 100 converts the constraint obtained in s10 into a constraint function as described above (s11). This constraint condition function is a function that is minimized when the constraint condition is satisfied. Note that when a constraint condition is specified in s10 described above, a constraint function defined for the constraint condition is extracted from the constraint table 126.

In addition, the schedule creation support device 100 uses the Ising model 1021 as an annealing machine, which has settings regarding the three items described above (the total number of working periods, the number of people required for each job on each day, and the constraint function). , calculate the ground state in which the objective function is minimized (s12). The search itself for such a ground state is similar to the processing in existing technology.

In other words, while satisfying the constraints on each operator and the stipulations of basic information such as the total number of shifts mentioned above and the number of people required for each job each day, each job of each day (theoretically based on sensitivity) is satisfied over time. The base state is a state in which the operator transitions toward the final state of the operating rate (taking skill level into account) and the result of the assigned operating rate of each operator settles down.

In addition, the schedule creation support device 100 transmits information on the assigned utilization rate of each operator in each time period specified in s12 (screen 900 in FIG. 9) as a schedule to the user terminal 200 (s13), and ends the process. .

On the other hand, the user terminal 200 that receives such information displays a schedule screen 900 as shown in FIG. 9 on an output unit such as a display.

The operator of the user terminal 200 views the schedule screen 900 and recognizes, for example, the timetable for next month, tomorrow, or the next hour, and appropriately confirms and considers the operation of the call center business.

In addition, as an Ising model that is simpler than the above example, an objective function that includes as terms the total number of working hours of each operator in a certain month, the number of people required for each job on each day of the month, and a function for constraint conditions. Regarding this, an Ising model may be assumed in which the operating rate of each job of each operator on each day is set as the spin, and the sensitivity between variables in the constraint function is set as the strength of the interaction between the spins.

In that case, an Ising model with one element, ie, skill level, excluded from spin in the form already described is assumed, and the ground state is searched for. Therefore, the calculation itself is simplified and the processing load is reduced. However, since the concept of proficiency does not apply to the accuracy of the resulting schedule, the accuracy also decreases. An example of a schedule obtained in this manner is shown in screen 1000 in FIG. 10.

In addition, as an Ising model that is simpler than the above example, each operator's objective function includes the total number of working hours of each operator in a certain month, the required number of people on each day of the month, and a constraint function. An Ising model may be assumed in which the daily operating rate of the operator is set as spin, and the sensitivity between variables in the constraint function is set as the strength of interaction between spins.

In that case, we will search for the ground state by assuming an Ising model in which one more element is excluded from the spin in the form already described, that is, not only proficiency but also utilization rate for each skill. Therefore, the calculation itself is further simplified and the processing load is reduced. However, the accuracy of the resulting schedule decreases because the concept of utilization rate for each skill is not considered in addition to proficiency level. An example of a schedule obtained in this manner is shown in screen 1100 in FIG. 11.

Furthermore, the schedule creation support device 100 is also capable of dealing with rare tasks that require important skills among call center tasks, for example. The rare task is, for example, a task that is rare enough to occur once a day, but can be assumed to be a task that requires delicate handling that cannot be handled lightly. Therefore, it is necessary to have people capable of handling these rare tasks available throughout the day, regardless of other tasks.

In this case, the schedule creation support device 100 determines the number of required personnel for the above-mentioned rare tasks, which are the constraint conditions, and the time period for responding to the rare tasks (e.g., the total working hours on a working day). Let us include the function in the objective function term and calculate the Ising model. The Ising model is based on an objective function that includes the total number of working hours of each operator in a certain month, the required number of people for each job on each day of the month, and a function for constraint conditions. An Ising model may be assumed in which the operating rate of a business (of course, rare jobs are included) is set as a spin, and the sensitivity between variables in a constraint function is set as the strength of interaction between spins.

Examples of schedules obtained in this manner are shown in screen 1200 in FIG. 12A and screen 1250 in FIG. 12B. In this embodiment, as shown in FIGS. 12A and 12B, a schedule is generated that takes only important skills (corresponding to rare work S) into consideration. In other words, in addition to the process of searching for the base state mentioned above for normal skills, for rare tasks corresponding to important skills, we also perform a The Ising model is used to explore its ground state and generate a schedule.

By generating such a schedule, it is possible to ensure that there is a staffing system that can reliably handle even rare tasks in parallel with regular tasks.

Although the above description has been about a configuration mainly based on the schedule creation support device 100, it is also possible to imagine an entire system including a CTI (Computer Telephony Integration) system 300 of a call center that is linked to such a configuration. FIG. 13 shows a configuration example of the CTI system 300 in this embodiment.

The CTI system 300 is a system that is generally operated as a call center management system, and is a platform that controls incoming and outgoing calls between operators and customers and links various computer resources and telephone resources on the call center side. .

Its general functions include, for example, a function that allocates inbound (incoming calls) calls to the terminal 400 of an operator who is not handling a customer and displays a screen for customer management when responding to the call from the customer. Equipped. Therefore, it is assumed that the CTI system 300 accumulates and manages information such as identification information of each operator to be managed and their terminals 400, customer information, and call history.

The CTI system 300 in this embodiment acquires the schedule generated by the schedule creation support device 100, and each operator regarding each job (that is, a job requiring a specific skill) in a certain time period indicated by the schedule. Automatically receive calls from customers during the relevant time period to the terminal 400 of each operator regarding each business (e.g., service application, service content change, service cancellation, etc.) according to the information on the operating rate of the operator. .

In this case, it is assumed that the CTI system 300 manages the length of telephone calls received regarding each business during the relevant time period on each operator's terminal 400. FIG. 14 shows an example of a management table 325 for managing this call time length.

In addition, the CTI system 300 calculates, based on the information on the call length in each time period managed as described above, the ratio of the call time length in the above time period for each operator to the utilization rate of the corresponding business determined in the above schedule. control to fit within the range.

For example, in a schedule, operator ) and a 10% break (that is, a total of 6 minutes).

In this case, as shown in the flowchart of FIG. 15, the CTI system 300 totals the call time on each phone for business A and business B that received calls to the terminal 400 of operator X for each call opportunity. , updates the information in the above-mentioned management table 325 (s20).

In addition, each time a customer calls, the CTI system 300 checks the schedule 310 (already obtained from the schedule generation support device 100 and stored in the storage unit 301) using the handling task and time of the call as keys. Then, in a time period including the current time, an operator corresponding to the task is specified (s21).

Next, for the operator identified in s21, the CTI system 300 refers to the call time length related to the business in the relevant time period in the above-mentioned management table 325, and calculates the call time length in the relevant time period (e.g. 1 hour). The ratio is calculated (s22). The value calculated here is the rate at which the operator worked on the job during the time period, that is, the current operating rate.

In addition, the CTI system 300 determines whether the current operating rate calculated in s22 does not exceed the operating rate of the business in the relevant time period specified for the operator in the schedule 310 described above (s23). .

As a result of the above determination, if the current operating rate does not exceed the specified operating rate in the schedule 310 (s23: OK), the CTI system 300 sends the call from the customer to the operator's terminal 400. A call is received (s24).

On the other hand, as a result of the above determination, if the current operating rate has reached the specified operating rate in the schedule 310 (s23: NG), the CTI system 300 is operated by another operator other than the operator, An operator whose current operating rate does not exceed the specified operating rate in the schedule 310 is identified (s25), and the above-mentioned call from the customer is received at the terminal 400 of the operator (s26).

For example, if the operating rate for business A reaches exactly 60% at a certain point in time, calls from customers regarding business A will not be received at the operator's terminal 400 during that time period, and business A will be terminated. The call is made to arrive at the terminal 400 of another operator who is in charge of the time period and whose operating rate has not reached the scheduled operating rate.

The CTI system 300 can perform operations that comply with the schedule 310 by continuously performing such control for each job in each time period, and can efficiently operate the call center appropriately.

Although the best mode for carrying out the present invention has been specifically described above, the present invention is not limited thereto and can be modified in various ways without departing from the gist thereof.

According to this embodiment, it is possible to efficiently create a schedule in which the necessary personnel are allocated for each task in each time period without waste.

The description of this specification clarifies at least the following. That is, in the schedule creation support device of the present embodiment, the storage unit stores information on the required number of people who may include decimal numbers at each timing for each required skill of each job in the period, and the skills possessed by each of the personnel. Further, the calculation unit stores: an objective function that includes as terms the total working length in the period, the required number of people for each required skill, and a constraint function that is minimized when the constraint conditions are satisfied; Calculate an Ising model in which the utilization rate for each skill of each person is set as spin, and the sensitivity between variables in the constraint condition function is set as the strength of interaction between the spins, and based on the result of the calculation, A schedule defining the operating rate of each of the personnel for each of the necessary skills at each of the timings of the predetermined period may be output to a predetermined device.

According to this, in the past, when assigned personnel were able to perform multiple skills (tasks), the ``proportion of performing each task'' was left to the discretion of the person or the on-site supervisor, and the optimal implementation ratio In order to deal with the issue of not being able to achieve this goal, it becomes possible to calculate in advance the optimal implementation ratio of skills (corresponding tasks) for each person and each time period (or day). This leads to more efficient personnel allocation and tends to allow each person to concentrate on their original work. In addition, it is expected that the current personnel will achieve the best performance, and operational efficiency can be improved. As a result, it becomes possible to more efficiently create a schedule that allocates the necessary personnel for each task in each time period without waste.

Further, in the schedule creation support device of the present embodiment, the calculation unit calculates the required number of personnel to handle the rare work and the time required to respond to the rare work, which are the constraint conditions regarding the rare work requiring important skills. Regarding the band, the Ising model may be calculated by including the constraint function in the objective function term.

According to this, in the past, schedulers would have to manually adjust situations such as ``one person with rare skills is always needed for a rare inquiry.''
By applying the concept of skill implementation ratio, it is possible to make settings such as ``always assigning one person with a rare skill'' without any special steps. As a result, it becomes possible to more efficiently create a schedule that allocates the necessary personnel for each task in each time period without waste.

Further, in the schedule creation support device of the present embodiment, the storage unit further stores information on the proficiency level of each skill possessed by each of the personnel, and the calculation unit stores the skill level of each of the personnel with respect to the objective function. An Ising model in which the operating rate for each spin is set as the spin, and the strength of the interaction between the spins is set as the sensitivity between variables in the constraint function multiplied by the proficiency level for each skill of each person. and, based on the result of the calculation, outputs to a predetermined device a schedule that defines the operating rate of each of the personnel by taking into account the proficiency level for each of the necessary skills at each timing of the predetermined period. , may also be used.

According to this, in the past, as long as the required number of people was an integer, even if calculations were made that took proficiency into account, the proficiency level of all personnel would end up being around 1 on average, so it would have little effect (or "new employees By reducing the number of people required and combining proficiency level with skill implementation rate/operation rate, it is possible to create shifts with fewer surpluses and shortages than before. (If a rookie comes in, the veteran's rest time can be increased, so the rest will be fair). As a result, it becomes possible to more efficiently create a schedule that allocates the necessary personnel for each task in each time period without waste.

Further, in the schedule creation support device of the present embodiment, the storage unit further stores information on the proficiency level of each skill possessed by each of the personnel, and the calculation unit stores the skill level of each of the personnel with respect to the objective function. An Ising model is calculated in which the operating rate and proficiency level for each spin is set as a spin, and the sensitivity between variables in the constraint function is set as the strength of interaction between the spins, and based on the result of the calculation, the predetermined A schedule may be output to a predetermined device in which the operating rate of each of the personnel is defined in consideration of the proficiency level for each of the necessary skills at each of the timings of the period.

According to this, in the past, as long as the required number of people was an integer, even if calculations were made that took proficiency into account, the proficiency level of all personnel would end up being around 1 on average, so it would have little effect (or "new employees By reducing the number of people required and combining proficiency level with skill implementation rate/operation rate, it is possible to create shifts with fewer surpluses and shortages than before. (If a rookie comes in, the veteran's rest time can be increased, so the rest will be fair). Moreover, by treating skill level in addition to spin, the skill level is not fixed, and it is possible to flexibly respond to changes such as improvement in skill level over time, or decrease in skill level due to updating of equipment, for example. As a result, it becomes possible to more efficiently create a schedule that allocates the necessary personnel for each task in each time period without waste.

Further, in the schedule creation support device of this embodiment, it may be a CMOS annealing machine that solves a combinatorial optimization problem regarding the Ising model.

According to this, the operation of the Ising model is simulated by a circuit using semiconductor elements such as CMOS (Complementary Metal Oxide Semiconductor), and combinations that take into account constraints that influence each other, that is, nonlinear constraints, are proposed. It becomes possible to efficiently find practical solutions to optimization problems at room temperature. As a result, it becomes possible to more efficiently create a schedule that allocates the necessary personnel for each task in each time period without waste.

10 Network 100 Schedule creation support device (annealing machine)
101 Storage unit 102 Program 1021 Ising model 103 Memory 104 Arithmetic unit 105 Communication unit 125 Basic information table 126 Constraint table 200 User terminal 300 CTI system 400 Operator terminal

Claims (7)

Stored information includes the total working length of each person who collaborates in a given task in a given period, the required number of people who may include decimal numbers at each timing in the period, and constraint conditions when assigning each person to the task. storage section and
Regarding the objective function that includes as terms the total working length in the period, the required number of people, and a constraint function that is the minimum when the constraint condition is satisfied, the utilization rate of each person is set as spin, and the constraint condition is a calculation unit that calculates an Ising model in which the sensitivity between variables in the function is set as the strength of interaction between the spins,
The calculation unit outputs to a predetermined device a schedule that defines the operating rate of each of the personnel at each timing of the predetermined period based on the result of the calculation.
A schedule creation support device characterized by: The storage unit includes:
further storing information on the required number of people, which may include decimal numbers, at each timing for each required skill of each job during the period, and the skills possessed by each of the personnel;
The arithmetic unit is
Regarding the objective function that includes as terms the total working length in the period, the required number of people for each required skill, and the minimum constraint function when the constraint is satisfied, the utilization rate for each skill of each of the personnel. is the spin, and calculates an Ising model in which the sensitivity between variables in the constraint function is set as the strength of the interaction between the spins,
Based on the result of the calculation, a schedule defining the operating rate of each of the personnel for each of the necessary skills at each timing of the predetermined period is output to a predetermined device;
The schedule creation support device according to claim 1, characterized in that: The arithmetic unit is
Regarding the constraint conditions regarding rare tasks that require important skills, including the required number of personnel to handle the rare tasks and the time period for responding to the rare tasks, the constraint function is included in the objective function section, which calculates the Ising model;
3. The schedule creation support device according to claim 2. The storage unit includes:
further retaining information on proficiency levels for each skill possessed by each of the personnel;
The arithmetic unit is
Regarding the objective function, the operating rate for each skill of each person is defined as a spin, and the sensitivity between variables in the constraint function is multiplied by the proficiency level for each skill of each person during the spin. Calculate the Ising model set as the interaction strength of
Based on the result of the calculation, a schedule is output to a predetermined device, in which the operating rate of each of the personnel is defined by taking into account the proficiency level for each of the necessary skills at each timing of the predetermined period;
3. The schedule creation support device according to claim 2. The storage unit includes:
further retaining information on proficiency levels for each skill possessed by each of the personnel;
The arithmetic unit is
Regarding the objective function, an Ising model is calculated in which the operating rate and proficiency level for each skill of each person is set as spin, and the sensitivity between variables in the constraint condition function is set as the strength of interaction between the spins. ,
Based on the result of the calculation, a schedule is output to a predetermined device, in which the operating rate of each of the personnel is defined by taking into account the proficiency level for each of the necessary skills at each timing of the predetermined period;
3. The schedule creation support device according to claim 2. 2. The schedule creation support device according to claim 1, wherein the schedule creation support device is a CMOS annealing machine that solves a combinatorial optimization problem regarding the Ising model. Stored information includes the total working length of each person who collaborates in a given task in a given period, the required number of people who may include decimal numbers at each timing in the period, and constraint conditions when assigning each person to the task. An information processing device including a storage unit,
Regarding the objective function that includes as terms the total working length in the period, the required number of people, and a constraint function that is the minimum when the constraint condition is satisfied, the utilization rate of each person is set as spin, and the constraint condition is calculate an Ising model in which the sensitivity between variables in the function is set as the strength of the interaction between the spins,
outputting a schedule defining the operating rate of each of the personnel at each timing of the predetermined period to a predetermined device based on the result of the calculation;
A schedule creation support method characterized by the following.

Images