Call center scheduling method based on heuristic genetic algorithm

Abstract

The application discloses a call center scheduling method based on a heuristic genetic algorithm and belongs to the technical field of operation optimization, and comprises the following steps: S1, inputting scheduling information; S2, determining daily work teams according to the scheduling information input in the step S1; S3, determining shifts corresponding to the daily work teams by using a heuristic genetic algorithm; and S4, obtaining a monthly scheduling roster according to the shifts corresponding to the daily work teams obtained. By using the heuristic algorithm, the application changes the traditional call center personnel scheduling mode, introduces the scheduling algorithm into the traditional scheduling, replaces the complicated manual scheduling work, realizes the monthly automatic scheduling of the call center personnel, can implement the intelligentization of the scheduling, reduces the time cost and the cost of the monthly scheduling of the call center personnel.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of operational optimization, in particular to a call center scheduling method based on heuristic genetic algorithm. BACKGROUND

[0002] A call center refers to an organization that provides services to enterprise customers through telephone by customer service representatives. With the rapid increase in the size of the call center, the skill classification is continuously refined, and the call center scheduling problem is becoming increasingly difficult and complex. Therefore, how to achieve the minimization of operating costs and the maximization of profits through efficient and economic scheduling, effectively guarantee the customer service level and service quality, and greatly improve the productivity of the call center, is an important problem in the operation and management of large call centers.

[0003] Specifically, the scheduling system analyzes the historical traffic data of the call center, scientifically predicts the traffic volume and the required staff in the future period, and then reasonably schedules according to the existing human resources of the call center. There are many methods for traffic volume prediction, such as support vector machine-based algorithm, neural network-based algorithm, and autoregressive-based algorithm. The Erlang-C formula or its improvement is used to obtain the predicted traffic volume to further calculate the predicted number of seats (half-hour manpower demand) every half hour.

[0004] The current call center scheduling mainly relies on the human scheduling of the scheduling teacher. The scheduling teacher establishes the predicted value of the conversation volume of each day in the next month according to the previous historical conversation data, then the scheduling teacher constructs an initial scheduling table, and finally the scheduling teacher adjusts the initial scheduling table according to the specific situation of each day to obtain the final call center personnel scheduling table; this method has the disadvantages of long time-consuming, large manpower demand, and long time span. The above problems need to be solved, and therefore a call center scheduling method based on heuristic genetic algorithm is proposed. SUMMARY

[0005] The technical problem to be solved by the present application is how to solve the problems of long time-consuming, large manpower demand, and long time span in the existing scheduling method, and a call center scheduling method based on heuristic genetic algorithm is provided. The method analyzes and fits the known half-hour manpower demand in the scheduling period, and finally realizes the matching of the corresponding shift for each day of the month for the shift team.

[0006] The present application solves the above technical problems by the following technical solutions, and the present application comprises the following steps:

[0007] S1: input scheduling information;

[0008] S2: determine the daily shift team according to the scheduling information input in step S1;

[0009] S3: determining the shift corresponding to each daily work team by using a heuristic genetic algorithm;

[0010] S4: obtaining the monthly scheduling roster according to the shift corresponding to each daily work team obtained.

[0011] Further, in the step S1, the scheduling information includes an initial scheduling roster, a team information table, and a shift information table; the initial scheduling roster contains information of total manpower supply per day and shift information of the initially scheduled team; the team information table includes team id and manpower of the team; and the shift information table includes shift id, name of the shift, and start and end time.

[0012] Further, the rest shift in the initial scheduling roster is fixed and cannot be changed in subsequent scheduling, and the initial scheduling roster is obtained according to the arrangement of a scheduling teacher.

[0013] Further, in the step S3, the existing half-hour manpower demand per day is fitted, that is, the number of seats is predicted to the greatest extent, so that the error between the number of people calculated at each half-hour point and the seat prediction result is minimized.

[0014] Further, the following constraints are met during fitting: each team can only perform one shift per day.

[0015] Further, the predicted traffic volume is obtained by a support vector machine algorithm, a neural network algorithm, or an autoregressive algorithm, and then the predicted number of seats per half hour, that is, the half-hour manpower demand, is calculated by the Erlang-C formula.

[0016] Further, in the step S3, the following process is specifically included:

[0017] S31: the scheduling problem can be abstracted into an optimization problem, and the following is obtained:

[0018] argminf(V) = MSE(Y, SUM_by_TIME(VWe X))

[0019]

[0020] Wherein, argminf(V) represents the value of variable V when the target f(V) takes the minimum value, MSE is the mean square error; "e" is the Hadamard product symbol; the SUM_BY_TIME(X) function is the sum of X according to time; Y is the overall demand; is a constraint condition, that is, the number of daily work teams is fixed;

[0021] There are M half hours in the total time period that needs to be scheduled per day, and the manpower demand number of the mth half hour is y mm = 1, 2, …, M, the overall demand vector is Y = [y1, y2, …, yM]. M ] T ; the input data includes N teams and K shifts, and the situation of each team performing shifts on the day is represented by matrix V:

[0022]

[0023]

[0024] The corresponding half hour of each shift is represented by matrix W:

[0025]

[0026]

[0027] The manpower supply of each team in the corresponding half hour is represented by X:

[0028]

[0029] x1, x2, …, x N is the number of manpower that team n can provide;

[0030]

[0031] S32: further obtain:

[0032]

[0033]

[0034] wherein y′ nm refers to the number of manpower that the nth team can provide in the mth half hour;

[0035] S33: finally obtain:

[0036] SUM_by_TIME(Y′) = [sum(Y′1), sum(Y′2), …, sum(Y′ M )];

[0037]

[0038] S34: calculate the matrix V using a heuristic genetic algorithm to obtain the corresponding shift of each daily working team.

[0039] Further, in the step S34, the heuristic algorithm includes any one of a particle swarm algorithm, a simulated annealing algorithm, a genetic algorithm, an ant colony optimization algorithm, and an artificial neural network algorithm.

[0040] Further, in the step S34, the heuristic genetic algorithm is a particle swarm algorithm, and the specific process is as follows:

[0041] S341: randomly initializing each particle, i.e. the shift of each team;

[0042] S342: evaluating each particle and obtaining a global optimum;

[0043] S343: judging whether the daily work and rest condition is met, if yes, ending the algorithm and outputting the shift corresponding to the team of daily work, otherwise, entering the step S344;

[0044] S344: updating the speed and position of each particle;

[0045] S345: evaluating the fitness value of each particle;

[0046] S346: updating the historical optimal position of each particle;

[0047] S347: updating the global optimal position of the group, entering the step S343, until the daily work and rest condition is met, and outputting the shift corresponding to the team of daily work.

[0048] Compared with the prior art, the call center scheduling method based on the heuristic genetic algorithm has the following advantages: the call center scheduling method based on the heuristic genetic algorithm changes the traditional call center personnel scheduling mode by using the heuristic algorithm, and the scheduling algorithm is introduced into the traditional scheduling to replace the complicated manual scheduling work, so that the monthly automatic scheduling of the call center personnel is realized, the intelligence of scheduling is implemented, the time cost of monthly scheduling of the call center personnel is reduced, and the cost is reduced. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 is a flowchart of the call center scheduling method based on the heuristic genetic algorithm in the embodiment one of the application;

[0050] Figure 2 is a process diagram for calculating the matrix V by using the heuristic genetic algorithm in the embodiment one of the application;

[0051] Fig. 3(a) is a first half of a line graph of the satisfaction of the manpower demand in the embodiment two of the application;

[0052] Fig. 3(b) is a second half of the line graph of the satisfaction of the manpower demand in the embodiment two of the application. DETAILED DESCRIPTION

[0053] The embodiments of the application will be described in detail below, and the embodiments are implemented on the premise of the technical scheme of the application, and detailed implementation modes and specific operation processes are given, but the protection scope of the application is not limited to the following embodiments.

[0054] Example 1

[0055] like Figure 1 As shown, this embodiment provides a technical solution: a call center scheduling method based on a heuristic genetic algorithm. This method is applied to the daily scheduling and includes the following steps:

[0056] Step 1: Enter shift information

[0057] In this step, the scheduling information includes the initial scheduling table, the work group information table, and the shift information table. The initial scheduling table contains information on the total daily manpower supply and the shift information for the initially scheduled work groups. The work group information table includes the work group ID and the number of people in the work group. The shift information table includes the shift ID, the name of the shift, and the start and end times.

[0058] Step 2: Determine the daily work shifts based on the shift information entered in Step 1.

[0059] In this step, considering that manual scheduling involves specific arrangements for the number and type of rest days (two days off or one day off per week, etc.) for each shift throughout the month, the scheduler needs a personalized initial schedule showing the distribution of rest shifts from a real business perspective. The rest shifts in the initial schedule will not be changed in subsequent scheduling; in other words, the initial schedule mainly provides information on the total daily manpower supply.

[0060] Step 3: Use a heuristic genetic algorithm to determine the shifts corresponding to the daily work groups.

[0061] In this step, the existing daily half-hour manpower demand is fitted, meaning the seat number prediction curve is fitted as closely as possible to minimize the error between the calculated number of people and the seat prediction result at each half-hour point. At the same time, certain constraints apply, such as each shift can only operate one shift per day.

[0062] The predicted call volume is obtained using algorithms based on support vector machines, neural networks, and autoregression. Then, the predicted number of agents per half hour, i.e., the manpower requirement per half hour, is calculated using Erlang-C formulas.

[0063] Specifically, this step includes the following specific procedures:

[0064] 1) The scheduling problem can be abstracted into an optimization problem, resulting in:

[0065] argminf(V)=MSE(Y,SUM_by_TIME(VWe X))

[0066]

[0067] Where argminf(V) represents the value of variable V that minimizes the objective f(V), MSE is the mean squared error; "e" is the Hadamard product symbol; the SUM_BY_TIME(X) function sums X over time; Y is the overall requirement; The constraint is that the number of work shifts is fixed each day;

[0068] The total daily shift schedule consists of M half-hour periods. The manpower requirement for the m-th half-hour is y. m The overall demand vector is Y = [y1, y2, ..., ym], where m = 1, 2, ..., M. M ] T The input data includes N work groups and K shifts. The daily shift schedule for each work group is represented by matrix V:

[0069]

[0070]

[0071] The half-hour corresponding to each shift is represented by matrix W:

[0072]

[0073]

[0074] The manpower provided by each shift in the corresponding half hour is represented by X:

[0075]

[0076] x1,x2,...,x N Let n be the number of workers that work group n can provide;

[0077]

[0078] 2) Thus we get:

[0079]

[0080]

[0081] Where, y′ nm This refers to the number of manpower that the nth shift can provide in the mth half-hour.

[0082] 3) The final result is:

[0083] SUM_by_TIME(Y')=[sum(Y′1),sum(Y′2),...,sum(Y′ M )];

[0084]

[0085] 4) Calculate the matrix V using the heuristic genetic algorithm (particle swarm algorithm) to obtain the corresponding shift of each daily working team, the specific process is as follows (see Figure 2 ) :

[0086] 41) Randomly initialize each particle, that is, the shift of each team;

[0087] 42) Evaluate each particle and obtain the global optimum;

[0088] 43) Determine whether the daily working and resting conditions are met, yes, end the algorithm, and output the corresponding shift of each daily working team, otherwise go to step 44) ;

[0089] 44) Update the speed and position of each particle;

[0090] 45) Evaluate the fitness value of each particle;

[0091] 46) Update the historical optimal position of each particle;

[0092] 47) Update the global optimal position of the group, go to step 43), until the daily working and resting conditions are met, and output the corresponding shift of each daily working team.

[0093] In step 4), the heuristic algorithm includes but is not limited to particle swarm algorithm, simulated annealing algorithm, genetic algorithm, ant colony optimization algorithm, artificial neural network algorithm.

[0094] Step 4: According to the corresponding shift of each daily working team obtained in step 3, the monthly scheduling table is obtained.

[0095] Example 2

[0096] Here, taking the scheduling of 10 teams in a certain district in August as an example, the scheduling shift type table and the monthly table finally scheduled by the method can be seen in Tables 1 and 2. The satisfaction of manpower demand can be seen in Figures 3(a) and (b), where the straight line “—” represents the number of manpower demand every half hour from traffic volume; the dashed line “--” is the corresponding manpower supply of the shift table formed by considering the real daily working team in step 3. It can be seen that the dashed line is generally below the straight line, indicating that the used team cannot fully meet the manpower demand, but the manpower supply corresponding to the scheduling table output by the method fluctuates with time, which is consistent with the manpower demand.

[0097] Table 1: Shift table

[0098]

[0099]

[0100] Table 2: Monthly shift table obtained by the method

[0101]

[0102]

[0103] Wherein the shift group id is obtained from the initial shift group, and is for identification purpose, and has no special meaning;

[0104] In summary, the call center scheduling method based on heuristic genetic algorithm of the above embodiments first determines the shift group working condition of each day in the month through relevant rules or regulations, and the date of each shift group working and resting in the month is determined; then, the specific shift that each shift group needs to complete is determined according to the half-hour manpower demand of each day using genetic algorithm, and the monthly scheduling shift table is given, which shows what shift each shift group corresponds to on each date in the month, and the rest corresponds to the "rest" shift; the human participation part of the whole process can be reduced to only inputting the initial scheduling table, shift group information and shift information, and completing automatic scheduling; the high adaptability is embodied in that when the number of shift group manpower and the demand are greatly deviated, in general shift allocation, all shift groups are allocated a few kinds of shifts, and through the form of shift group dynamic round-robin allocation of shifts, the types of shifts allocated by the shift group are more and cover all time periods in a day.

[0105] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.

Claims

1. A call center scheduling method based on a heuristic genetic algorithm, characterized in that, Includes the following steps: S1: Enter shift schedule information; S2: Determine the daily work shifts based on the scheduling information entered in step S1; S3: Use a heuristic genetic algorithm to determine the shifts corresponding to the daily work groups; S4: Based on the daily work shifts, the monthly work schedule is obtained. In step S3, specifically Includes the following processes: S31: The scheduling problem can be abstracted into an optimization problem, resulting in: arg min f(V)=MSE(Y,SUM_by_TIME(VW⊙X)) Where arg min f(V) represents finding the value of variable V that minimizes the objective f(V), MSE is the mean squared error; "⊙" is the Hadamard product symbol; the SUM_BY_TIME(X) function sums X over time; Y is the overall requirement; The constraint is that the number of work shifts is fixed each day; The total daily shift schedule consists of M half-hour periods. The manpower requirement for the m-th half-hour is y. m The overall demand vector is Y = [y1, y2, ..., ym], where m = 1, 2, ..., M. M ] T The input data includes N work groups and K shifts. The daily shift schedule for each work group is represented by matrix V: The half-hour corresponding to each shift is represented by matrix W: The manpower provided by each shift in the corresponding half hour is represented by X: x1,x2,...,x N Let n be the number of workers that work group n can provide; S32: Therefore, we get: Where, y′ nm This refers to the number of manpower that the nth shift can provide in the mth half-hour. S33: The final result is: SUM_by_TIME(Y′)=[sum(Y1′),sum(Y′2),...,sum(Y′ M )]; S34: Calculate matrix V using a heuristic genetic algorithm to obtain the shifts corresponding to the daily work groups.

2. The call center scheduling method based on a heuristic genetic algorithm according to claim 1, characterized in that: In step S1, the scheduling information includes an initial scheduling table, a team information table, and a shift information table. The initial scheduling table includes information on the total daily manpower supply and the shift information for the initially scheduled teams. The team information table includes the team ID and the number of people in the team. The shift information table includes the shift ID, the name of the shift, and the start and end times.

3. The call center scheduling method based on a heuristic genetic algorithm according to claim 2, characterized in that: The rest shifts in the initial schedule will remain fixed and will not be changed in subsequent schedules. The initial schedule is obtained according to the scheduler's arrangement.

4. The call center scheduling method based on a heuristic genetic algorithm according to claim 3, characterized in that: In step S3, the existing half-hour manpower demand for each day is fitted, that is, the seat number prediction curve is fitted to the maximum extent, so that the error between the number of people calculated at each half-hour point and the seat prediction result is minimized.

5. A call center scheduling method based on a heuristic genetic algorithm according to claim 4, characterized in that: The fitting process is subject to the following constraint: each shift can only perform one shift per day.

6. A call center scheduling method based on a heuristic genetic algorithm according to claim 5, characterized in that: The predicted call volume is obtained using algorithms based on support vector machines, neural networks, or autoregression. Then, the predicted number of agents per half hour, i.e., the manpower requirement per half hour, is calculated using Erlang-C formulas.

7. The call center scheduling method based on a heuristic genetic algorithm according to claim 1, characterized in that: In step S34, the heuristic algorithm includes any one of particle swarm optimization, simulated annealing, genetic algorithm, ant colony optimization, or artificial neural network algorithm.

8. A call center scheduling method based on a heuristic genetic algorithm according to claim 7, characterized in that: In step S34, the heuristic genetic algorithm is a particle swarm optimization algorithm, and the specific process is as follows: S341: Randomly initialize each particle, i.e., the shift in each group; S342: Evaluate each particle and obtain the global optimum; S343: Determine whether the conditions for daily work and rest are met. If yes, end the algorithm and output the shift number corresponding to the daily work group. Otherwise, proceed to step S344. S344: Update the velocity and position of each particle; S345: Evaluate the functional fitness value of each particle; S346: Update the best historical position for each particle; S347: Update the global optimal position of the group, proceed to step S343, until the conditions for daily work and rest are met, and output the shifts corresponding to the daily work groups.

Images