Method, apparatus, device, and medium for processing shift scheduling scheme based on business volume

Abstract

An embodiment of the present invention discloses a method, device, equipment, and medium for processing a scheduling plan based on business volume. The method includes: obtaining first historical business volume data with a set time period granularity; inputting the first historical business volume data into a pre-trained business volume prediction model to predict and determine the predicted business volume with the set time period granularity within a set time range; calculating a target scheduling manpower requirement plan according to the predicted business volume of each time period within the time range to be scheduled, the set service capacity value, and the target service index value; calculating a candidate scheduling plan based on the existing manpower data and scheduling hard constraint conditions using an operations research optimization algorithm; calculating the matching result between personnel and business according to the candidate scheduling plan and the target scheduling manpower requirement plan, and iteratively adjusting the candidate scheduling plan according to the matching result and the target optimization function until the conditions of the target optimization function are met. The present invention achieves the effect of improving the rationality and efficiency of the scheduling plan.

Description

Technical Field

[0001] Embodiments of the present invention relate to artificial intelligence technology, machine learning technology, and big data operations research optimization technology, and particularly to a scheduling plan processing method, device, equipment, and medium based on business volume. Background Art

[0002] On the operator side, it is often necessary for operation staff to provide various business services to users online, such as interacting via telephone or instant messaging to answer questions for users or handle business matters.

[0003] The functions in the existing scheduling management module include operation staff information management, shift-related functions, manual scheduling, and shift schedule adjustment-related functions. Existing scheduling mainly relies on the experience of management personnel, lacks prediction based on business volume, as well as corresponding human resource conversion and automatic scheduling. Moreover, with the continuous changes in business scale, service types, and personnel qualities, the factors tend to become more complex, making the scheduling complexity increasingly high. However, the automation degree of existing scheduling operations is relatively low, which is cumbersome and time-consuming, and it is difficult to meet the requirements of multiple parties. Summary of the Invention

[0004] Embodiments of the present invention provide a scheduling plan processing method, device, equipment, and medium based on business volume to fully consider the impact of business volume on the scheduling plan and improve the rationality and efficiency of the scheduling plan.

[0005] In a first aspect, embodiments of the present invention provide a scheduling plan processing method based on business volume, including:

[0006] Obtain first historical business volume data with a set time period granularity;

[0007] Input the first historical business volume data into a pre-trained business volume prediction model to predict and determine the predicted business volume with a set time period granularity within a set time range; wherein, the business volume prediction model is a model based on the regression tree algorithm, the attributes of the first historical business volume data include at least one time impact factor and an associated time impact factor, and the first historical business volume data includes actual business volume data and statistical business volume data generated based on the actual business volume data;

[0008] Calculate a target scheduling human resource demand plan according to the predicted business volume of each time period within the time range to be scheduled, a set service capacity value, and a target service index value;

[0009] Calculate a candidate scheduling plan based on the existing human resource data and scheduling hard constraint conditions using an operations research optimization algorithm;

[0010] According to the candidate shift scheduling plan and the target shift scheduling manpower requirement plan, calculate the matching result of personnel and business, and iteratively adjust the candidate shift scheduling plan according to the matching result and the target optimization function until the conditions of the target optimization function are met, where the target optimization function includes shift soft constraint conditions.

[0011] In a second aspect, an embodiment of the present invention further provides a shift scheduling plan processing device based on business volume, including:

[0012] A first historical business volume data acquisition module, configured to acquire first historical business volume data with a set time period granularity;

[0013] A business volume prediction module, configured to input the first historical business volume data into a pre-trained business volume prediction model to predict and determine the predicted business volume with a set time period granularity within a set time range; wherein, the business volume prediction model is a model based on a regression tree algorithm, and the attributes of the first historical business volume data include at least one time impact factor and an associated moment impact factor, and the first historical business volume data includes actual business volume data and statistical business volume data generated based on the actual business volume data;

[0014] A target shift scheduling calculation module, configured to calculate a target shift scheduling manpower requirement plan according to the predicted business volume of each time period within the time range to be scheduled, a set service capacity value, and a target service index value;

[0015] A candidate shift scheduling calculation plan, configured to calculate a candidate shift scheduling plan based on the existing manpower data and shift hard constraint conditions using an operations research optimization algorithm;

[0016] A matching and optimization module, configured to calculate the matching result of personnel and business according to the candidate shift scheduling plan and the target shift scheduling manpower requirement plan, and iteratively adjust the candidate shift scheduling plan according to the matching result and the target optimization function until the conditions of the target optimization function are met, where the target optimization function includes shift soft constraint conditions.

[0017] In a third aspect, an embodiment of the present invention further provides a computer device, including:

[0018] One or more processors;

[0019] A memory, configured to store one or more programs,

[0020] When the one or more programs are executed by the one or more processors, the one or more processors implement the shift scheduling plan processing method based on business volume as described in any one of the embodiments.

[0021] Fourthly, an embodiment of the present invention further provides a storage medium containing computer-executable instructions, which are used to execute the traffic-based shift scheduling scheme processing method as described in any one of the embodiments when executed by a computer processor.

[0022] The technical solution of this embodiment is to obtain historical traffic data with a set time period granularity; input the historical traffic data into a pre-trained traffic prediction model to predict and determine the predicted traffic with the set time period granularity within a set time range; wherein, the traffic prediction model is a model based on the regression tree algorithm, and the attributes of the historical traffic data include at least one time impact factor and an associated time impact factor, and the historical traffic data includes actual traffic data and statistical traffic data generated based on the actual traffic data; calculate the target shift scheduling manpower requirement plan according to the predicted traffic of each time period within the time range to be scheduled, the set service capacity value, and the target service index value; calculate a candidate shift scheduling plan based on the operation research optimization algorithm according to the existing manpower data and the hard shift scheduling constraints; calculate the matching result between personnel and business according to the candidate shift scheduling plan and the target shift scheduling manpower requirement plan, and iteratively adjust the candidate shift scheduling plan according to the matching result and the target optimization function until the conditions of the target optimization function are met, wherein the target optimization function includes soft shift scheduling constraints, which solves the problem that the existing shift scheduling strongly depends on the experience of managers, has a low degree of automation, is cumbersome and time-consuming, and is difficult to meet the requirements of multiple parties, and achieves the effect of improving the rationality and efficiency of the shift scheduling plan. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a flowchart of a traffic-based shift scheduling scheme processing method provided in Embodiment 1 of the present invention;

[0024] Figure 2a is a schematic diagram of the monthly traffic regular line in a bank;

[0025] Figure 2b is a schematic diagram of the weekly traffic regular line in a bank;

[0026] Figure 2c is a schematic diagram of the daily traffic regular line in a bank;

[0027] Figure 3 is a flowchart of a traffic-based shift scheduling scheme processing method provided in Embodiment 2 of the present invention;

[0028] Figure 4 is a structural diagram of a traffic-based shift scheduling scheme processing device provided in Embodiment 3 of the present invention;

[0029] Figure 5 is a schematic structural diagram of a device provided in Embodiment 4 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0030] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.

[0031] The technical solution of the embodiment of the present invention aims to automatically allocate the working hours and personnel allocation of the operation staff. For example, a bank branch has a total of 10 customer service staff. At the end of March, it is necessary to schedule the customer service work in April to determine which day and time period in April these 10 customer service staff will work. The above example is only the simplest case. In fact, when scheduling the operation staff, it is also necessary to consider various factors such as individual differences and work ability of employees. For example, it is also necessary to consider allocating more operation staff during peak business hours. That is to say, when determining the scheduling plan, two aspects of influence are usually considered. One is the business volume to be scheduled, such as the size of the business volume and the type of business, and the other is the situation of the operation staff to be scheduled. Of course, other influencing factors will also be considered. From the perspective of mathematical programming theory, the automatic allocation process is essentially a 0-1 integer programming problem. During the establishment process, the decision variables are binary variables of 0 and 1. For example, 0 indicates that a certain operation staff works in a certain time period, and 1 indicates that the operation staff does not work in this time period. Among them, in the process of modeling and solving, linear constraints are used as much as possible to facilitate solving.

[0032] Taking into account the periodicity, stability, and trend of business volume distribution, as well as the impact of business activities on the business, the business volume forecast adopts a machine learning algorithm. By constructing feature engineering data and combining the impact of business activities on the business, it meets the requirements of forecast accuracy, stability, and fit in different months.

[0033] Embodiment 1

[0034] Figure 1 This is a flowchart of a method for processing a shift scheduling scheme based on business volume provided in Embodiment 1 of the present invention. This embodiment is applicable to situations where automatic scheduling is required. The method can be executed by a device for processing a shift scheduling scheme based on business volume. The device can be implemented in software and / or hardware. The device can be configured in a computer device with computing capabilities, such as a server or other device. The method specifically includes the following steps:

[0035] S110: Acquire first historical business volume data at a set time period granularity.

[0036] In this embodiment, to determine the final shift scheduling plan, the business volume of the shifts to be scheduled is first considered. Since the shifts to be scheduled will occur in a future period of time, it is necessary to predict the business volume of the shifts to be scheduled by analyzing the business volume data of a past period of time. The first historical business volume data refers to the business volume data of a past period of time used to predict the business volume of the shifts to be scheduled. Among them, the time period granularity refers to the smallest unit of time when obtaining the first historical business volume data. For example, when obtaining the first historical business volume data of the past three years and setting the time period granularity to 15 minutes, the first historical business volume data of the past three years is divided based on 15 minutes. One hour can be divided into 4 first historical business volume data blocks, and the first historical business volume data of the past three years is obtained with many 15-minute first historical business volume data blocks. It should be noted that the time granularity is set arbitrarily according to the situation. In the present invention, since the shift scheduling cycle is 30 minutes, that is, a new batch of operating personnel is changed every 30 minutes, a time granularity less than 30 minutes is selected. If the shift scheduling cycle changes, the time granularity can also be changed accordingly. For example, the time granularity is 30 minutes. However, for the accuracy of the prediction results, the time granularity should be less than the shift scheduling cycle.

[0037] S120. Input the first historical business volume data into a pre-trained business volume prediction model to predict and determine the predicted business volume with a set time period granularity within a set time range.

[0038] Input the first historical business volume data with the set time period granularity into the business volume prediction model to obtain the predicted business volume, that is, obtain the business volume of the shifts to be scheduled.

[0039] Among them, the business volume prediction model is a model based on the regression tree algorithm. The attributes of the first historical business volume data include at least one time influence factor and an associated moment influence factor. The first historical business volume data includes actual business volume data and statistical business volume data generated based on the actual business volume data.

[0040] In the first historical traffic data, two types of data can be defined in the first historical traffic data. One type is the traffic data for a whole period of time before the period to be predicted. For example, if the traffic for a certain 15 minutes in the next month is to be predicted, all the traffic data for the past three years needs to be obtained, and the attributes of the traffic for every 15 minutes in the past three years constitute the time impact factor. The second type is the traffic data for a specific period before the period to be predicted. For example, if the traffic for a certain 15 minutes is to be predicted, the traffic data for the corresponding 15 minutes on the previous day needs to be obtained, and the attributes of the traffic for these corresponding 15 minutes constitute the associated moment impact factor. It should be noted that the associated moment impact factor actually belongs to a part of the time impact factor. However, since the associated moment impact factor has a greater impact on the prediction result, the associated moment impact factor is extracted from the time impact factor and given more attention. The selection of the associated moment impact factor can also be based on the principle of larger value for nearer time. For example, the traffic for the past thirty days closest to the prediction time period can be statistically analyzed, and the average value can be used as the associated moment impact factor.

[0041] Optionally, the time impact factor may include at least one of the following: month attribute, date attribute, weekday attribute, holiday attribute, week attribute, number of weekdays in the current month attribute, billing date attribute, and SMS date attribute. The associated moment of the associated moment impact factor may include at least one of the following: the same moment on the day before the prediction moment, the same moment on the two days before the prediction moment, and the same moment on the set date in the first three weeks before the prediction moment.

[0042] Continuing with the previous example, taking a first historical traffic data block of 15 minutes from the historical traffic data of the past three years as an example, its time impact factor can include multiple tags to reflect the attributes of this 15-minute historical traffic data block, which can be January 1st, a working day, Wednesday, the number of working days in the current month is 24 days, a non-billing day, and a non-text message day respectively. At the same time, when its associated moment is the same moment of the previous day of the prediction moment, the attributes of the historical traffic data block corresponding to the 15 minutes of the previous day of the 15 minutes to be predicted also need to be extracted, which can be December 31st, a working day, Tuesday, the number of working days in the current month is 25 days, a non-billing day, and a non-text message day. When using a computer for training, the above attributes can be marked as numerical values for the convenience of computer recognition. For example, December is represented by the number 12, a working day is represented by the number 1, and a holiday is represented by the number 0. For example, when the distribution characteristics of the traffic volume show that the business handling is concentrated at the beginning and end of the month, which is the peak period, while the traffic volume on other days is relatively small, then in this case, the time impact factor should include the date attribute, and this situation is also called that the monthly regular line is obvious. It should be noted that the selection of the associated moment can be set according to the situation, and the moment with the strongest correlation is selected, not limited to the same moment of the previous day, the previous two days, or the previous three weeks, but can also be other moments with a relatively strong correlation with the moment to be predicted.

[0043] The advantage of such a setting is that by abstracting the complex data into simple numerical values and using attribute values to represent the characteristics of the moment, it is convenient for computer calculation and recognition.

[0044] In the process of designing the embodiments of the present invention, the time distribution characteristics of the traffic volume in the bank are statistically analyzed. Figure 2a It is a schematic diagram of the monthly regular line of the traffic volume in the bank. Figure 2a Each curve in it represents the traffic volume situation of each month. Among them, the statistical data includes multiple Januarys, multiple Februarys, …, multiple Novembers, and multiple Decembers. Although the traffic volume situations of each month are not the same, the trend is the same, that is, the traffic volume is larger at the beginning and end of the month, and smaller in the middle of the month. Figure 2b It is a schematic diagram of the weekly regular line of the traffic volume in the bank. Figure 2b Each curve in it represents the traffic volume situation of each day in a week. Among them, the statistical data includes multiple Mondays, multiple Tuesdays, …, multiple Saturdays, and multiple Sundays. Although the traffic volume situations of each day are not the same, the trend is the same, that is, the traffic volume is larger in the morning and afternoon of the day, and smaller at noon, when just opening, and when about to get off work. Figure 2c It is a schematic diagram of the daily regular line of the traffic volume in the bank. Figure 2cEach curve in it represents the business volume of each day. Among them, although the business volume situation is different every day, the trend is the same, that is, the business volume is relatively large from 9:00 to 10:00 in the morning and from 2:00 to 4:00 in the afternoon, and the business volume is relatively small at other times.

[0045] The first historical business volume data includes actual business volume data and statistical business volume data generated based on the actual business volume data. The actual business volume data refers to the obtained original business volume data, while the statistical business volume data refers to the business volume data obtained after processing such as extraction, secondary processing, or classification and averaging.

[0046] Optionally, the statistical business volume data generated based on the actual business volume data may include: the mean calculated based on the actual business volume data within a set recent time range, as the statistical business volume.

[0047] For example, when predicting the business volume at a certain moment in units of 15 minutes in May of the future, the first historical business volume data of the past three years obtained is the original business volume data, and the statistical business volume data is obtained by averaging the business volume data of April, which is the closest to May in the past three years.

[0048] The advantage of such a setting is that by setting the statistical business volume data, the business volume data with more reference value in the actual business volume data is used as the key in the business volume prediction model, thereby improving the accuracy of the prediction result.

[0049] It should be noted that the first historical business volume data is time series data. According to the time attribute characteristics of the time series data, in addition to using the above-mentioned machine learning algorithm for prediction, it can also be analyzed and predicted based on traditional statistical exponential smoothing and ARIMA models.

[0050] S130. Calculate the target shift scheduling man - power requirement plan according to the predicted business volume of each period within the time range of the shift to be scheduled, the set service capacity value, and the target service index value.

[0051] The predicted business volume value has been obtained in S120. That is to say, the business volume situation to be scheduled has been obtained. In S130, the situation of the operation staff to be allocated is further determined. The set service capacity here refers to the employee ability profile, such as the attendance rate of employees, the average processing time of business, and the experience value, etc. That is to say, due to the different service capabilities of the operation staff, the same business volume may require different numbers of operation staff to complete. Therefore, when calculating the candidate scheduling plan, the individual differences of the staff need to be considered. The target service index value can be understood as the expected target service level. For example, when the target service index value is 0.8, it means that under the condition that the business volume to be scheduled is 10 items, completing 8 items is qualified. That is to say, the higher the target service index value, the more operation staff are required to complete the same business volume. Therefore, when calculating the candidate scheduling plan, the target service index value also needs to be considered. The calculated target scheduling manpower requirement plan can be, for example: under the condition of completing 10 items of business in the predicted 15 minutes, 3 full-time operation staff or 5 interns are required. Among them, the target scheduling manpower requirement plan can be calculated based on the Erlang-C (international standard call center management software) formula.

[0052] S140. Calculate the candidate scheduling plan based on the existing manpower data and scheduling hard constraint conditions using the operation research optimization algorithm.

[0053] The operation research optimization algorithm is a general term for algorithms, which refers to various algorithms used in the process of obtaining the required results based on known conditions. In this embodiment, it means using the existing manpower data and scheduling hard constraint conditions as conditions and the operation research optimization algorithm as a tool to obtain the candidate scheduling plan as the result.

[0054] Optionally, the existing human resource data includes the number of personnel in each unit, the unit to which the personnel belong, the business position skill value of the personnel, the experience ability weight value of the personnel, and shift information. Specifically for each employee, the human resource data may include: the center where the employee is located, the group, the employee number, the employee name, the position, the role, the start time of the work schedule, the end time of the work schedule, the time when the work schedule is released, and the situation where there are different skills among the scheduled personnel in the same position. When scheduling, it is necessary to schedule separately according to different positions and different skills of the scheduled personnel. Among them, the experience ability weight value of the employee can be understood as a characteristic in the employee ability portrait. For example, the experience ability weight value of an old employee can be 1.2, indicating that an old employee is equivalent to 1.2 ordinary full-time operation staff, while the experience ability weight value of an intern is 0.8, indicating that an intern is equivalent to 0.8 ordinary full-time operation staff. The shift information indicates the types of shifts that an operation staff can undertake. For example, an operation staff can undertake two types of shifts, the morning shift and the evening shift, while another staff can only undertake the morning shift. The business position skill value of the personnel refers to the types of business that a staff is good at or cannot engage in. If it is predicted that there will be a large number of specific types of business tomorrow, then select the staff who is good at this type of business to undertake the work tomorrow. The advantage of such a setting is that it pays attention to the individual differences of the operation staff, making the candidate scheduling plan more reasonable.

[0055] The scheduling hard constraints refer to certain conditions that must be met during the scheduling process. Optionally, the scheduling hard constraints include at least one of the following: the number of rest days, working hours constraints, consecutive working days, rotation prohibition, and shift prohibition. For example, an operation staff must have two days off per week, or each person can be scheduled for a maximum of 6 hours per day, cannot work continuously for more than 7 days, must rest after the evening shift, and a certain operation staff cannot be scheduled for the evening shift, etc. That is to say, it is necessary to sort out the hard rules during scheduling, such as including the shift cycle, shifts, meal times, and special needs of caring personnel. The advantage of such a setting is that before obtaining the final scheduling result, the conditions that must be met are determined first to ensure the feasibility of the final obtained scheduling result.

[0056] It should be noted that the target scheduling human resource demand plan and the candidate scheduling plan are obtained in S130 and S140 respectively. These two plans are obtained according to different conditions and are two different plans. Among them, the target scheduling human resource demand plan is obtained based on the volume of business to be scheduled, while the candidate scheduling plan is obtained based on the situation of the operation staff to be scheduled.

[0057] S150. Calculate the matching result of personnel and business according to the candidate scheduling plan and the target scheduling human resource demand plan, and iteratively adjust the candidate scheduling plan according to the matching result and the target optimization function until the conditions of the target optimization function are met.

[0058] Among them, the target optimization function includes shift scheduling soft constraint conditions. Optionally, the shift scheduling soft constraint conditions include at least one of the following: shift fairness conditions, shift comfort priority order, and shift relationships of personnel from the same unit. For example, the shift fairness conditions may refer to whether the number of morning shifts or the number of evening shifts worked by each employee in a month is equal; the shift comfort priority order may refer to sorting the comfort levels of shifts and preferentially allocating business periods with higher comfort levels under the same conditions; the shift relationships of personnel from the same unit means that personnel from the same unit can be preferentially allocated to work during the same period, making management more convenient. The advantage of such a setting is that it provides an optimization direction for the shift scheduling plan for the business volume, further satisfying the shift scheduling soft constraint conditions while meeting the shift scheduling hard constraint conditions, making the shift scheduling plan more reasonable.

[0059] The shift comfort priority order may further include: the shift comfort priority order is Day 1 shift > Day 2 shift > morning shift > evening shift, where the priorities are sorted from high to low.

[0060] The shift fairness conditions may further include: the difference in the number of evening shifts of any two operating staff within the same scheduling cycle does not exceed one day. Assuming this value is X and the difference number is 1, the difference in the shift schedule results can only be X + 1 or X - 1, and the situations of X + 1 and X - 1 cannot exist simultaneously. For example: in the shift schedule for the scheduling cycle in July 2021, when setting the difference in the number of rest days to 1, when arranging the total number of rest days of a certain operating staff within the attendance cycle to be 5, the rest days of any two operating staff can only be 5 days, 6 days or 4 days, 5 days, and no other array situations can exist; within the same scheduling cycle, the fairness of the morning shifts, Day 2 shifts, and evening shifts requires that the total number of shifts arranged at the beginning and end of the month be relatively fair. For example: in the July shift schedule, if 10 evening shifts are arranged at the beginning of the month, about 10 evening shifts also need to be arranged at the end of the month.

[0061] S150 is essentially a process of matching the target shift scheduling manpower demand plan and the candidate shift scheduling plan. Since the candidate shift scheduling plan and the target shift scheduling manpower demand plan are obtained based on different conditions, the presented plans are different. The result obtained from the target shift scheduling manpower demand plan is virtual personnel, while the candidate shift scheduling plan is specific to real people, respectively reflecting the requirements of both the supply and demand sides of manpower and business for the shift scheduling plan.

[0062] Since the business requirements cannot be changed, the object of optimization can only be the candidate shift scheduling plan, that is, iteratively adjusting the candidate shift scheduling plan according to the matching result and the target optimization function to obtain the optimal shift scheduling plan.

[0063] The technical solution of this embodiment is to obtain historical traffic data with a set time period granularity; input the historical traffic data into a pre-trained traffic prediction model to predict and determine the predicted traffic with the set time period granularity within a set time range. Among them, the traffic prediction model is a model based on the regression tree algorithm. The attributes of the historical traffic data include at least one time impact factor and an associated time impact factor. The historical traffic data includes actual traffic data and statistical traffic data generated based on the actual traffic data; calculate the target shift scheduling manpower requirement plan according to the predicted traffic of each time period within the time range to be scheduled, the set service capacity value, and the target service index value; calculate the candidate shift scheduling plan based on the existing manpower data and shift scheduling hard constraints using the operations research optimization algorithm; calculate the matching result of personnel and business according to the candidate shift scheduling plan and the target shift scheduling manpower requirement plan, and iteratively adjust the candidate shift scheduling plan according to the matching result and the target optimization function until the conditions of the target optimization function are met. Among them, the target optimization function includes shift scheduling soft constraints, which solves the problem that the existing shift scheduling strongly depends on the experience of managers, has low automation, is cumbersome and time-consuming, and is difficult to meet the requirements of multiple parties, and achieves the effect of improving the rationality and efficiency of the shift scheduling plan.

[0064] Embodiment 2

[0065] Figure 3 FIG. is a flowchart of a traffic-based shift scheduling plan processing method provided by the second embodiment of the present invention. This embodiment is applicable to the situation where automatic shift scheduling is required and is a further refinement of Embodiment 1. The specific steps are as follows. The parts that are the same as those in Embodiment 1 will not be described again:

[0066] Train the traffic prediction model, which specifically includes S310 and S320.

[0067] The process of training the traffic prediction model is the process of modeling according to historical data, specifically including: remote authorization, post-event supervision, corporate credit review, account review, early warning monitoring, foreign exchange settlement, capital business inspection, business consultation, credit operation analysis, special business, management assistance, retail business inspection, corporate and capital business inspection, retail loan review, credit operation inspection, centralized entry position, centralized review position, etc. for traffic volume, each system interface, business time accounting, and total traffic volume verification.

[0068] S310. Obtain the second historical traffic data with a set time period granularity, and determine the traffic data of the historical moment and the corresponding predicted moment from the second historical traffic data as a training sample pair.

[0069] The second historical traffic volume data here is different from the above-mentioned first historical traffic volume data and refers to the historical traffic volume data used to train the traffic volume prediction model. Continuing with the example in Embodiment 1, since the first historical traffic volume data for the past three years is used to predict the next 15 minutes during prediction, therefore, when training the traffic volume prediction model, the second historical traffic volume data for the past three years and the traffic volume data for the subsequent 15 minutes are also extracted as a training sample pair. For example, when the historical time in the second historical traffic volume data is from 0:00 to 0:15 on August 1, 2019, the traffic volume data corresponding to the prediction time is from 0:00 on August 1, 2016 to 24:00 on July 31, 2019; when the historical time in the second historical traffic volume data is from 0:00 to 0:15 on June 25, 2015, the traffic volume data corresponding to the prediction time is from 0:00 on June 25, 2012 to 24:00 on June 24, 2015. The data volume of the second historical traffic volume data is usually larger than that of the first historical traffic volume data, and multiple training sample pairs are extracted from the second historical traffic volume data.

[0070] S320. Input the training sample pair into the traffic volume prediction model for training.

[0071] Input the large number of training sample pairs obtained in S310 into the traffic volume prediction model, where the traffic volume prediction model is a model based on the regression tree algorithm. The regression tree algorithm is a commonly used algorithm in the field of machine learning and is usually used to construct a prediction model.

[0072] Optionally, the regression tree algorithm can be the XGBOOST algorithm. XGBOOST is a machine learning algorithm based on decision trees. It efficiently implements the Gradient Boosting Decision Tree (GBDT) algorithm and has made many improvements in algorithms and engineering. XGBOOST is widely used in the industrial and academic fields and has achieved good results. XGBOOST is essentially the Gradient Boosting Decision Tree (GBDT) algorithm. The Gradient Boosting Decision Tree is a supervised learning algorithm, which is an additive combination of a series of Classification and Regression Trees (CART). The subsequent tree fits the "residual" between the previous prediction result and the target.

[0073] The XGBoost algorithm has the following characteristics: regularization feature. XGBoost adds a regularization term to the cost function to control the complexity of the model, prevent overfitting, and thus improve the generalization ability of the model; missing value handling feature. XGBoost treats missing values as sparse matrices and does not consider the numerical values of missing values during node classification. For samples with missing feature values, XGBoost places the sample in the left subtree and the right subtree respectively, calculates the gain respectively, and selects the better one; built-in cross-validation function feature. The XGBoost algorithm performs cross-validation during each iteration; approximate algorithm feature. When splitting tree nodes, we need to calculate the gain corresponding to each split point of each feature, that is, use the greedy algorithm to enumerate all possible split points. When the data cannot be loaded into memory at one time or in a distributed scenario, the efficiency of the greedy algorithm becomes very low. Therefore, XGBoost also proposes a parallelizable approximate algorithm for efficiently generating candidate split points.

[0074] When using the XGBoost algorithm to build a prediction model, it is necessary to extract the time attribute features and value features of the first historical traffic volume data and the second historical traffic volume data. After multiple rounds of data training, prediction, and verification, important feature values are obtained, thereby effectively improving the prediction accuracy.

[0075] The traffic volume prediction model obtained through training must be tested before it can be put into use. That is to say, the prediction result obtained using the traffic volume prediction model must meet the accuracy requirements when compared with the actual traffic volume in order to be effectively used in human conversion and scheduling models. Generally, when the average business prediction accuracy reaches 65%, the allowable prediction deviation is between -5% and 10%, and the maximum system prediction deviation is less than 10%, it is considered that the traffic volume prediction model meets the requirements.

[0076] The traffic volume prediction model obtained through training is tested through goodness-of-fit tests. Goodness-of-fit tests refer to comparing the degree of agreement between the prediction results of the traffic volume prediction model and the actual occurrence. Usually, several prediction models are tested simultaneously, and the one with better goodness-of-fit is selected for trial use. Common goodness-of-fit test methods include: sum of squared residuals test, chi-square (c2) test, and linear regression test, etc. Among them, the statistic for measuring goodness-of-fit is the coefficient of determination (also known as the determination coefficient) R 2 , R 2 The maximum value is 1. The closer the value of R 2 is to 1, the better the fitting degree; conversely, the smaller the value of R 2 , the worse the fitting degree.

[0077] S330. Obtain the first historical traffic volume data with a set time period granularity.

[0078] S340. Determine the data missing period in the first historical business volume data, and use the nearest neighbor node algorithm to generate the business volume data for the missing period based on the business volume data of the adjacent periods of the missing period, and fill it into the missing period of the first historical business volume data.

[0079] That is to say, when the first historical business volume data within a certain period is missing and is 0, the business volume data of the adjacent periods can be used to fill it. For example, when the data of a certain 15 minutes in the first historical business volume data is missing, the average value of the business volume data of the two adjacent 15 minutes before and after it can be used for filling. Of course, other filling methods can also be adopted.

[0080] Specifically, a box plot can be used to identify the abnormally missing parts in the first historical business volume data, and the identified missing parts are replaced, and the KNN (K-Nearest Neighbor) algorithm is used to fill the missing parts in the first historical business volume data.

[0081] It should be noted that although only the technical solution for filling the first historical business volume data is described here, it is easy to think that in the case of missing second historical business volume data, the same or similar method can also be used to fill the missing parts in the second historical business volume data.

[0082] S350. Distinguish the first historical business volume data according to business positions, and input the first historical business volume data of at least two business positions into the pre-trained business volume prediction models corresponding to at least two business positions respectively, so as to predict and determine the predicted business volume of each business position at the set time granularity within the set time range.

[0083] That is to say, in S350, the first historical business volume data is further subdivided according to business positions. Correspondingly, when training the business volume prediction model, the second historical business volume data should also be subdivided according to the same standard, so as to train business volume prediction models for different business positions. For example, a business volume model for consulting services and a business volume model for handling services are obtained.

[0084] That is to say, according to the first historical business volume data and the corresponding production capacity data, the time series data is cleaned, analyzed and predicted to obtain the data for prediction that distinguishes business types and positions.

[0085] The advantage of such a setting is that by obtaining the business volume predictions of each different business position, operation staff who are good at different business types can be targeted to match and provide services for customers.

[0086] S360. Calculate the target scheduling manpower requirement plan according to the predicted business volume, set service capacity value, and target service index value for each time period within the time range to be scheduled. The target scheduling manpower requirement plan includes the required number of personnel for each time period and each business position.

[0087] Since the predicted business volume is differentiated according to business positions in this embodiment, the target scheduling manpower requirement plan is correspondingly divided according to business positions.

[0088] S370. Calculate the candidate scheduling plan based on the existing manpower data and scheduling hard constraint conditions using an operations research optimization algorithm. The candidate scheduling plan can also be divided according to business positions.

[0089] S380. Calculate the matching result of personnel and business according to the candidate scheduling plan and the target scheduling manpower requirement plan, and iteratively adjust the candidate scheduling plan according to the matching result and the target optimization function until the conditions of the target optimization function are met.

[0090] Optionally, calculating the matching result of personnel and business according to the candidate scheduling plan and the target scheduling manpower requirement plan, and iteratively adjusting the candidate scheduling plan according to the matching result and the target optimization function until the conditions of the target optimization function are met includes: converting the candidate scheduling plan and the target scheduling manpower requirement plan into a candidate scheduling curve and a target scheduling manpower requirement curve; calculating the approximate distance value between the candidate scheduling curve and the target scheduling manpower requirement curve according to the target optimization function; when the approximate distance value does not reach the maximization condition, iteratively adjust the candidate scheduling plan until the approximate distance value reaches the maximization condition.

[0091] That is to say, represent the candidate scheduling plan and the target scheduling manpower requirement plan as two curves respectively. Among them, the target scheduling manpower requirement plan cannot be changed. Therefore, it is necessary to iteratively adjust and change the candidate scheduling plan to make the candidate scheduling plan as close as possible to the target scheduling manpower requirement plan. The target optimization function here is established according to the scheduling soft constraint conditions. The maximization in the condition that the approximate distance value reaches the maximization means that the similarity between the candidate scheduling curve and the target scheduling manpower requirement curve reaches the maximum, and the distance value between the two curves reaches the minimum. The method for judging that the distance value between the two curves reaches the minimum can be that after several consecutive iterative adjustments of the candidate scheduling plan, the distance value does not decrease further, then it is considered that the approximate distance value has reached the maximization condition.

[0092] That is to say, through the target optimization function, a certain number of business personnel with different skills are arranged at the most appropriate working hours as much as possible, and a shift scheduling plan within a certain time period (for example: one month) is formulated while taking into account the comfort and fairness requirements of employees. That is, to maximize the utilization rate of employees, or to maximize the overall revenue, while meeting the requirements of shift manpower fitting, as well as the fairness and comfort constraints of employees.

[0093] The advantage of such a setting is that it provides a specific method for optimizing the candidate shift scheduling plan, and can maximize the satisfaction of the soft shift scheduling constraints under the condition of meeting the hard shift scheduling constraints.

[0094] The following describes in detail the process of iteratively adjusting the candidate shift scheduling plan according to the matching result and the target optimization function until the conditions of the target optimization function are met:

[0095] Since the shift scheduling rules include fairness rules and comfort rules, the automatic shift scheduling uses python+mathprog to implement the linear programming solution algorithm. According to the manpower requirements at each time point, the working hours of shifts, and the personnel situation participating in the shift scheduling, combined with each shift scheduling rule, the shift schedule of each employee within a shift scheduling period is obtained. The algorithm is mainly divided into the following steps:

[0096] 1. Convert each shift scheduling rule into a set of linear constraints and find the optimal solution of the objective function.

[0097] 2. Use the symmetric simplex algorithm to perform elimination and obtain a feasible solution.

[0098] 3. Use the gradient descent algorithm to find the optimal solution.

[0099] The process of establishing the automatic shift scheduling objective function is not to solve for the lowest cost or the maximum revenue, but to try to meet the requirements by maximizing the following values:

[0100] 1) The maximum value of the intermediate auxiliary variable (the weight can be set from 100 to 1000 as needed) to meet the fitting of the requirements and achieve the expected solution.

[0101] 2) The maximum satisfaction of the rotation rule (weight 100-1000) to implement the rotation rule. Since it is only a weight, there can be a breakthrough and non-satisfaction, but there must be a solution. In other words, as long as there is a solution space judged, there must be a solution.

[0102] 3) Strong rules are prohibited

[0103] In the embodiments of the present invention, a linear programming-based shift scheduling algorithm can be implemented using Python + MathProg. Linear rules are established for hard constraints, and soft constraints and connection metrics are placed in the objective function, thus achieving multi-objective constraint optimization. The symmetric simplex method is used for global optimization to obtain the global optimal solution. In the comparative experiment with the genetic algorithm for solving, it is closer to the actual business requirements of the centralized processing center.

[0104] Hard constraints, such as linear rules like y = ax + b, which can be two-dimensional, with a small number of three-dimensional ones, etc., can be directly implemented in the code for calculating the shift scheduling plan and are constraint conditions that must be met. Soft constraints, on the other hand, can be placed in the objective function and do not have to be satisfied, but are preferably satisfied.

[0105] The technical solution of this embodiment determines the data missing periods in the first historical traffic volume data, and uses the nearest neighbor node algorithm to generate the traffic volume data for the missing periods based on the traffic volume data of the adjacent periods of the missing periods, and fills it into the missing periods of the first historical traffic volume data; the first historical traffic volume data is distinguished according to business positions, and the first historical traffic volume data of at least two business positions are respectively input into the traffic volume prediction models corresponding to at least two business positions that have been pre-trained to predict and determine the predicted traffic volume of each business position at a set time range with a set time period granularity; and the candidate shift scheduling plan and the target shift scheduling manpower requirement plan are converted into a candidate shift scheduling curve and a target shift scheduling manpower requirement curve, and according to the target optimization function, the approximate distance value between the candidate shift scheduling curve and the target shift scheduling manpower requirement curve is calculated; when the approximate distance value does not reach the maximization condition, the candidate shift scheduling plan is iteratively adjusted until the approximate distance value reaches the maximization condition, solving the problems of inaccurate traffic volume prediction in the case of missing historical traffic volume data and how to specifically adjust the candidate shift scheduling plan in practice, and achieving the effect of further improving the rationality of the shift scheduling plan.

[0106] Embodiment III

[0107] Figure 4 It is a structural diagram of a traffic volume-based shift scheduling plan processing device provided in Embodiment III of the present invention. This device can execute a traffic volume-based shift scheduling plan processing method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0108] A traffic volume-based shift scheduling plan processing device includes:

[0109] A first historical traffic volume data acquisition module 410, configured to acquire first historical traffic volume data with a set time period granularity;

[0110] The traffic volume prediction module 420 is used to input the first historical traffic volume data into a pre-trained traffic volume prediction model to predict and determine the predicted traffic volume within a set time range at a set time granularity; wherein, the traffic volume prediction model is a model based on the regression tree algorithm, and the attributes of the first historical traffic volume data include at least one time impact factor and associated time impact factor, and the first historical traffic volume data includes actual traffic volume data and statistical traffic volume data generated based on the actual traffic volume data;

[0111] The target scheduling calculation module 430 is used to calculate the target scheduling manpower requirement plan according to the predicted traffic volume of each time period within the time range to be scheduled, the set service capacity value, and the target service index value;

[0112] The candidate scheduling calculation scheme 440 is used to calculate the candidate scheduling scheme based on the existing manpower data and scheduling hard constraints by means of an operations research optimization algorithm;

[0113] The matching and optimization module 450 is used to calculate the matching result between personnel and business according to the candidate scheduling scheme and the target scheduling manpower requirement plan, and iteratively adjust the candidate scheduling scheme according to the matching result and the target optimization function until the conditions of the target optimization function are met, wherein the target optimization function includes scheduling soft constraints.

[0114] Optionally, the traffic volume-based scheduling scheme processing device further includes:

[0115] The traffic volume prediction model training module is used to obtain the second historical traffic volume data at a set time granularity, and determine the traffic volume data at the historical time and the corresponding predicted time from the second historical traffic volume data as a training sample pair;

[0116] Input the training sample pair into the traffic volume prediction model for training.

[0117] Optionally, the time impact factor includes at least one of the following: month attribute, date attribute, weekday attribute, holiday attribute, week attribute, number of weekdays in the current month attribute, billing date attribute, and SMS date attribute.

[0118] Optionally, the associated time of the associated time impact factor includes at least one of the following: the same time of the day before the predicted time, the same time of the two days before the predicted time, and the same time of the set date in the three weeks before the predicted time.

[0119] Optionally, the statistical traffic volume data generated based on the actual traffic volume data includes: the mean value calculated based on the actual traffic volume data within a set recent time range as the statistical traffic volume.

[0120] Optionally, the traffic volume-based scheduling scheme processing device further includes:

[0121] A missing filling module is used to determine the data missing period in the first historical business volume data, and adopt the nearest neighbor node algorithm to generate the business volume data for the missing period based on the business volume data of the adjacent periods of the missing period, and fill it into the missing period of the first historical business volume data.

[0122] Optionally, the regression tree algorithm is the XGBOOST algorithm.

[0123] Optionally, the business volume prediction module 420 includes:

[0124] A business position prediction sub-module is used to distinguish the first historical business volume data according to business positions, and input the first historical business volume data of at least two business positions into the pre-trained business volume prediction models corresponding to at least two business positions respectively to predict and determine the predicted business volume of each business position at a set time granularity within a set time range.

[0125] Optionally, the target shift scheduling manpower requirement plan includes the required number of people for each period and each business position.

[0126] Optionally, the existing manpower data includes the number of personnel in each unit, the unit to which the personnel belong, the business position skill value of the personnel, the experience ability weight value of the personnel, and the shift information.

[0127] Optionally, the shift scheduling hard constraint conditions include at least one of the following: the number of rest days, working hours constraint, consecutive working days, rotation prohibition, and shift prohibition.

[0128] Optionally, the shift scheduling soft constraint conditions include at least one of the following: shift fairness condition, shift comfort priority order, and shift relationship of personnel in the same unit.

[0129] Optionally, the matching optimization module 450 includes:

[0130] A curve conversion sub-module is used to convert the candidate shift scheduling plan and the target shift scheduling manpower requirement plan into a candidate shift scheduling curve and a target shift scheduling manpower requirement curve;

[0131] A distance value calculation sub-module is used to calculate the approximate distance value between the candidate shift scheduling curve and the target shift scheduling manpower requirement curve according to the target optimization function;

[0132] An iterative adjustment sub-module is used to iteratively adjust the candidate shift scheduling plan until the approximate distance value reaches the maximization condition when the approximate distance value does not reach the maximization condition.

[0133] The technical solution of this embodiment is to obtain historical traffic data with a set time period granularity; input the historical traffic data into a pre-trained traffic prediction model to predict and determine the predicted traffic with the set time period granularity within a set time range; wherein, the traffic prediction model is a model based on the regression tree algorithm, and the attributes of the historical traffic data include at least one time impact factor and an associated time impact factor, and the historical traffic data includes actual traffic data and statistical traffic data generated based on the actual traffic data; calculate a target shift scheduling manpower requirement plan according to the predicted traffic of each time period within the time range to be scheduled, the set service capacity value, and the target service index value; calculate a candidate shift scheduling plan based on the operation research optimization algorithm according to the existing manpower data and shift scheduling hard constraints; calculate the matching result of personnel and business according to the candidate shift scheduling plan and the target shift scheduling manpower requirement plan, and iteratively adjust the candidate shift scheduling plan according to the matching result and the target optimization function until the conditions of the target optimization function are met, wherein the target optimization function includes shift scheduling soft constraints, which solves the problems that the existing shift scheduling strongly depends on the experience of managers, has a low degree of automation, is cumbersome and time-consuming, and is difficult to meet the requirements of multiple parties, and achieves the effect of improving the rationality and efficiency of the shift scheduling plan.

[0134] Embodiment 4

[0135] Figure 5 FIG. is a schematic structural diagram of a computer device provided in Embodiment 4 of the present invention, as Figure 5 shown, the device includes a processor 520, a memory 510, an input device 530, and an output device 540; the number of processors 520 in the device can be one or more, Figure 5 taking one processor 520 as an example; the processor 520, the memory 510, the input device 530, and the output device 540 in the device can be connected through a bus or other means, Figure 5 taking the connection through the bus as an example.

[0136] The memory 510, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules, such as program instructions / modules corresponding to the topic update method in the embodiments of the present invention (for example, the first historical traffic data acquisition module 410, the traffic prediction module 420, the target shift scheduling calculation module 430, the candidate shift scheduling calculation plan 440, and the matching optimization module 450 in a traffic-based shift scheduling plan processing device). The processor 520 executes various functional applications and data processing of the device by running the software programs, instructions, and modules stored in the memory 510, that is, implements the above-mentioned traffic-based shift scheduling plan processing method.

[0137] The memory 510 may mainly include a program storage area and a data storage area. Among them, the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created according to the use of the terminal, etc. In addition, the memory 510 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage devices. In some instances, the memory 510 may further include a memory remotely provided with respect to the processor 520, and these remote memories may be connected to the device through a network. Examples of the above network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0138] The input device 530 may be used to receive input digital or character information, and generate key signal inputs related to the user settings and function controls of the device. The output device 540 may include a display device such as a display screen.

[0139] Embodiment 5

[0140] Embodiment 5 of the present invention further provides a storage medium containing computer-executable instructions, and the computer-executable instructions are used to execute a search method when executed by a computer processor. The method includes:

[0141] Obtain first historical traffic data with a set time period granularity;

[0142] Input the first historical traffic data into a pre-trained traffic prediction model to predict and determine the predicted traffic with a set time period granularity within a set time range; wherein, the traffic prediction model is a model based on a regression tree algorithm, and the attributes of the first historical traffic data include at least one time impact factor and an associated time impact factor, and the first historical traffic data includes actual traffic data and statistical traffic data generated based on the actual traffic data;

[0143] Calculate a target scheduling man-hour requirement plan according to the predicted traffic of each time period within the time range to be scheduled, the set service capacity value, and the target service index value;

[0144] Calculate a candidate scheduling plan based on the existing manpower data and scheduling hard constraint conditions based on an operations research optimization algorithm;

[0145] Calculate the matching result of personnel and business according to the candidate scheduling plan and the target scheduling man-hour requirement plan, and iteratively adjust the candidate scheduling plan according to the matching result and the target optimization function until the conditions of the target optimization function are met, wherein the target optimization function includes scheduling soft constraint conditions.

[0146] Of course, for a storage medium containing computer-executable instructions provided by an embodiment of the present invention, the computer-executable instructions are not limited to the method operations described above, and can also execute relevant operations in the traffic volume-based shift scheduling scheme processing method provided by any embodiment of the present invention.

[0147] From the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented by means of software and necessary general-purpose hardware. Of course, it can also be implemented by hardware, but in many cases the former is a better implementation. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as a floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk, or optical disc of a computer, and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments of the present invention.

[0148] It should be noted that in the embodiments of the above search device, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of mutual distinction and do not limit the protection scope of the present invention.

[0149] Note that the above is only a preferred embodiment of the present invention and the applied technical principle. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, re-adjustments, and substitutions can be made by those skilled in the art without departing from the protection scope of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments. Without departing from the concept of the present invention, more other equivalent embodiments can be included, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A scheduling plan processing method based on business volume, characterized in that, Including: Obtain the first historical traffic data with a set time period granularity; Input the first historical traffic data into a pre-trained traffic prediction model to predict and determine the predicted traffic with a set time period granularity within a set time range; wherein, the traffic prediction model is a model based on the regression tree algorithm, and the attributes of the first historical traffic data include at least one time impact factor and an associated moment impact factor. The time impact factor is used to reflect the attributes of the traffic volume of the first historical traffic data, and the associated moment impact factor is used to reflect the traffic volume attributes corresponding to the time period with the strongest correlation with the set time range determined based on all the time impact factors. The first historical traffic data includes actual traffic data and statistical traffic data generated after data preprocessing of the actual traffic data; Calculate the target scheduling manpower requirement plan according to the predicted traffic volume of each time period within the time range to be scheduled, the set service capacity value, and the target service index value; Calculate a candidate scheduling plan based on the existing manpower data and scheduling hard constraints using an operations research optimization algorithm, where the scheduling hard constraints refer to the conditions that must be met during the scheduling process; Calculate the matching result of personnel and business according to the candidate scheduling plan and the target scheduling manpower requirement plan, and iteratively adjust the candidate scheduling plan according to the matching result and the target optimization function until the conditions of the target optimization function are met. The target optimization function includes scheduling soft constraints, and the scheduling soft constraints include shift fairness conditions, shift comfort priority order, and shift relationships of personnel in the same unit.

2. The method according to claim 1, characterized in that, It also includes the training process of the traffic prediction model, and the training process includes: Obtain the second historical traffic data with a set time period granularity, and determine the traffic data of the historical moment and the corresponding prediction moment from the second historical traffic data as a training sample pair; Input the training sample pair into the traffic prediction model for training.

3. The method according to claim 1, wherein The time impact factor includes at least one of the following: month attribute, date attribute, weekday attribute, holiday attribute, week attribute, number of weekdays in the current month attribute, billing date attribute, and SMS date attribute.

4. The method according to claim 1, wherein The associated moments of the associated moment impact factor include at least one of the following: the same moment of the day before the prediction moment, the same moment of the two days before the prediction moment, and the same moment of the set date in the three weeks before the prediction moment.

5. The method according to claim 1, wherein, The statistical traffic data generated based on the actual traffic data includes: the mean calculated based on the actual traffic data within a set recent time range as the statistical traffic volume.

6. The method according to claim 1 or 2, characterized in that, After obtaining the first historical traffic data with a set time period granularity, it also includes: Determine the data missing time periods in the first historical traffic data, and use the nearest neighbor node algorithm to generate the traffic data of the missing time periods according to the traffic data of the adjacent time periods of the missing time periods, and fill them into the missing time periods of the first historical traffic data.

7. The method according to claim 1 or 2, characterized in that, The regression tree algorithm is the XGBOOST algorithm.

8. The method according to claim 1 or 2, characterized in that, Inputting the first historical traffic volume data into a pre-trained traffic volume prediction model to predict and determine the predicted traffic volume at a set time granularity within a set time range includes: Classify the first historical traffic volume data according to business positions, and input the first historical traffic volume data of at least two business positions into the pre-trained traffic volume prediction models corresponding to at least two business positions respectively, so as to predict and determine the predicted traffic volume of each business position at the set time granularity within the set time range.

9. The method according to claim 1, characterized in that: The target shift scheduling manning requirement plan includes the required number of personnel for each time period and each business position.

10. The method according to claim 1, characterized in that: The existing manning data includes the number of personnel in each unit, the unit to which the personnel belong, the business position skill value of the personnel, the experience ability weight value of the personnel, and the shift information.

11. The method according to claim 1, characterized in that: The hard shift scheduling constraints include at least one of the following: number of rest days, working hour constraints, consecutive working days, rotation prohibition, and shift prohibition.

12. The method according to claim 1, characterized in that: The soft shift scheduling constraints include at least one of the following: shift fairness conditions, shift comfort priority order, and shift relationship of personnel in the same unit.

13. The method according to claim 1, wherein Calculating the matching result of personnel and business according to the candidate shift scheduling plan and the target shift scheduling manning requirement plan, and iteratively adjusting the candidate shift scheduling plan according to the matching result and the target optimization function until the conditions of the target optimization function are met, including: Converting the candidate shift scheduling plan and the target shift scheduling manning requirement plan into a candidate shift scheduling curve and a target shift scheduling manning requirement curve; Calculating the approximate distance value between the candidate shift scheduling curve and the target shift scheduling manning requirement curve according to the target optimization function; When the approximate distance value does not reach the maximization condition, iteratively adjust the candidate shift scheduling plan until the approximate distance value reaches the maximization condition.

14. A scheduling plan processing device based on traffic volume, characterized in that, Including: A first historical traffic volume data acquisition module, configured to acquire the first historical traffic volume data at a set time granularity; A traffic volume prediction module, configured to input the first historical traffic volume data into a pre-trained traffic volume prediction model to predict and determine the predicted traffic volume at a set time granularity within a set time range; wherein, the traffic volume prediction model is a model based on a regression tree algorithm, and the attributes of the first historical traffic volume data include at least one time impact factor and an associated moment impact factor. The time impact factor is used to reflect the attributes of the traffic volume of the first historical traffic volume data, and the associated moment impact factor is used to reflect the traffic volume attributes corresponding to the time period with the strongest correlation with the set time range determined based on all the time impact factors. The first historical traffic volume data includes actual traffic volume data and statistical traffic volume data generated after data preprocessing of the actual traffic volume data; A target shift scheduling calculation module, configured to calculate a target shift scheduling manning requirement plan according to the predicted traffic volume of each time period within the time range to be scheduled, the set service capacity value, and the target service index value; A candidate shift scheduling calculation plan, configured to calculate a candidate shift scheduling plan based on the existing manning data and the hard shift scheduling constraints using an operational research optimization algorithm, where the hard shift scheduling constraints refer to the conditions that must be met during the shift scheduling process; A matching optimization module, configured to calculate a matching result between personnel and services according to the candidate shift scheduling plan and the target shift scheduling manpower requirement plan, and iteratively adjust the candidate shift scheduling plan according to the matching result and a target optimization function until the conditions of the target optimization function are met, wherein the target optimization function includes shift soft constraint conditions, and the shift soft constraint conditions include shift fairness conditions, shift comfort priority orders, and shift relationships of personnel in the same unit.

15. A computer device, characterized in that, The device includes: One or more processors; A memory for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method for processing a shift scheduling plan based on business volume according to any one of claims 1-13.

16. A storage medium containing computer-executable instructions, which are used to execute the method for processing a shift scheduling plan based on business volume according to any one of claims 1-13 when executed by a computer processor.

Images