Labor demand prediction method and device, and electronic equipment

Abstract

The application provides a labor demand prediction method, which comprises the following steps: constructing a time series decomposition model according to a plurality of historical total sales data corresponding to a plurality of historical time periods respectively; generating a plurality of predicted total sales data corresponding to a plurality of prediction time periods respectively based on the time series decomposition model; for each predicted total sales data, dividing the predicted total sales data into a plurality of predicted sub-sales data corresponding to each prediction sub-time period respectively based on the predicted total sales data and a plurality of average ratios obtained in advance; and generating labor demand information corresponding to each prediction sub-time period respectively based on each predicted sub-sales data. In this way, the predicted total sales data of the plurality of prediction time periods generated by the time series decomposition model can be divided according to the respective average ratios, the future sales can be predicted, the personnel scheduling can be reasonably performed, the rationality of the scheduling can be improved, and personnel redundancy or insufficiency can be avoided.

Description

Technical Field

[0001] This invention relates to the technical field of labor demand, and in particular to a method, apparatus and electronic device for predicting labor demand. Background Technology

[0002] Cashiers are an important part of supermarket and retail store operations. Cashier performance is generally based on the number of items scanned at the checkout, which is characterized by quantifiable workload.

[0003] Currently, the scheduling of cashiers in supermarkets and stores is mainly done manually by store managers and department heads. This scheduling efficiency is low, and the accuracy and rationality of the scheduling are not based on data. It relies heavily on the personal experience of the scheduling staff and does not take into account actual workload changes and employee work efficiency. This can easily lead to low scheduling rationality, resulting in redundancy or shortage of cashiers on certain dates or time periods. Redundancy increases the ineffective working hours of cashiers, while shortage affects the customer experience. Summary of the Invention

[0004] The purpose of this invention is to provide a method, apparatus, and electronic equipment for predicting labor demand, so as to improve the rationality of scheduling and avoid redundancy or shortage of personnel.

[0005] This invention provides a method for predicting labor demand, the method comprising:

[0006] A time series decomposition model is constructed based on multiple historical total sales data corresponding to multiple historical time periods obtained in advance.

[0007] Based on the time series decomposition model, multiple predicted total sales data corresponding to multiple prediction time periods are generated.

[0008] For each predicted total sales data, based on the predicted total sales data and multiple pre-acquired average ratios, the predicted total sales data is divided into multiple predicted sub-sales data corresponding to each predicted sub-time period.

[0009] Based on the sales data for each forecast sub-segment, generate the labor demand information corresponding to each forecast sub-time period.

[0010] Furthermore, each historical time period includes multiple historical sub-time periods; for each historical time period, the total historical sales data corresponding to that historical time period is the sum of the historical sub-sales data corresponding to each historical sub-time period of that historical time period; each average ratio is calculated in the following way:

[0011] Obtain the historical sub-sales data corresponding to each historical sub-time period;

[0012] Based on each historical sub-sales data and each historical total sales data, the proportion of historical sub-sales data corresponding to each historical sub-time period is obtained;

[0013] Based on the proportion of each historical sub-sales data, determine the average proportion corresponding to each historical sub-time period.

[0014] Furthermore, the steps for generating labor demand information corresponding to each forecast sub-period based on the sales data of each forecast sub-sub-period include:

[0015] Obtain the target historical smart purchase sales ratio for each historical sub-time period;

[0016] Based on the sales data of each predicted sub-sales segment and the historical intelligent purchase sales ratio of each target, determine the manual bank sales data corresponding to each predicted sub-time period.

[0017] Based on the sales data of each ICBC branch, the labor demand information corresponding to each predicted sub-period is generated.

[0018] Furthermore, the steps to obtain the target historical smart purchase sales ratio for each historical sub-time period include:

[0019] Obtain the historical intelligent purchase sales data corresponding to each historical sub-time period;

[0020] Based on each historical smart purchase sales data and each historical sub-sales data, the historical smart purchase sales ratio corresponding to each historical sub-time period is obtained;

[0021] Based on the sales volume ratio of each historical smart purchase, obtain the target historical smart purchase sales volume ratio corresponding to each historical sub-time period.

[0022] Furthermore, based on the sales data of each predicted sub-segment and the historical intelligent purchase sales ratio of each target, the steps to determine the manual bank sales data corresponding to each predicted sub-period include:

[0023] Based on the sales data of each predicted sub-segment and the historical smart purchase sales ratio of each target, the predicted smart purchase sales data corresponding to each predicted sub-period is determined.

[0024] Calculate the difference between each predicted sub-sales data and each predicted intelligent purchase sales data to obtain the manual bank sales data corresponding to each predicted sub-time period.

[0025] Furthermore, the methods also include:

[0026] Based on pre-set evaluation indicators, the rationality of the employment demand information corresponding to each predicted sub-period is evaluated.

[0027] Furthermore, the evaluation indicators include the average number of orders processed per target unit of working hours and the target coefficient of variation; based on the pre-set evaluation indicators, the steps for evaluating the rationality of the labor demand information corresponding to each predicted sub-period include:

[0028] Obtain the labor demand information corresponding to each forecast sub-period and the actual labor sales data corresponding to each forecast period;

[0029] Based on the labor demand information corresponding to each prediction sub-time period, obtain the total labor demand information corresponding to each prediction time period.

[0030] Based on the actual sales data of each manual bank counter and the total labor demand information, determine the order volume per unit working hour for each forecast period.

[0031] Based on the order volume per unit working hour corresponding to each forecast time period, determine the average order volume per unit working hour.

[0032] The standard deviation of the order volume processed per unit of working hours is determined based on the mean of the order volume processed per unit of working hours and the order volume processed per unit of working hours for each forecast period.

[0033] The coefficient of variation is obtained by calculating the ratio of the standard deviation of the number of orders processed per unit of working hours to the mean of the number of orders processed per unit of working hours.

[0034] If the average number of orders processed per unit of working hours is greater than the target average number of orders processed per unit of working hours and the coefficient of variation is less than the target coefficient of variation, then the employment demand information is considered reasonable.

[0035] This invention provides a labor demand forecasting device, the device comprising:

[0036] The building module is used to construct a time series decomposition model based on multiple historical total sales data corresponding to multiple historical time periods obtained in advance.

[0037] The first generation module is used to generate multiple predicted total sales data corresponding to multiple predicted time periods based on the time series decomposition model.

[0038] The segmentation module is used to divide each predicted total sales data into multiple predicted sub-sales data corresponding to each predicted sub-time period, based on the predicted total sales data and multiple pre-acquired average ratios.

[0039] The second generation module is used to generate employment demand information corresponding to each forecast sub-time period based on the sales data of each forecast sub-sub-period.

[0040] The present invention provides an electronic device, which includes a processor and a memory. The memory stores computer-executable instructions that can be executed by the processor, and the processor executes the computer-executable instructions to implement any of the methods described above.

[0041] The present invention provides a computer-readable storage medium storing computer-executable instructions, which, when invoked and executed by a processor, cause the processor to implement any of the methods described above.

[0042] The aforementioned method for forecasting labor demand includes: constructing a time series decomposition model based on multiple historical total sales data corresponding to multiple historical time periods obtained in advance; generating multiple predicted total sales data corresponding to multiple prediction time periods based on the time series decomposition model; for each predicted total sales data, dividing the predicted total sales data into multiple predicted sub-sales data corresponding to each prediction sub-time period based on the predicted total sales data and multiple pre-obtained average proportions; and generating labor demand information corresponding to each prediction sub-sales data. This method, by dividing the predicted total sales data of multiple prediction time periods generated by the time series decomposition model according to their respective average proportions, can predict future sales, thereby enabling reasonable personnel scheduling, improving the rationality of scheduling, and avoiding personnel redundancy or shortage. Attached Figure Description

[0043] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0044] Figure 1 A flowchart of a labor demand forecasting method provided in an embodiment of the present invention;

[0045] Figure 2 A schematic diagram illustrating the correlation between store labor efficiency and store size, sales volume, and shift work hours, provided as an embodiment of the present invention;

[0046] Figure 3 A flowchart of another labor demand forecasting method provided in an embodiment of the present invention;

[0047] Figure 4 This invention provides a schematic diagram illustrating the sales volume during a specific time period and the sales volume ratio during a smart purchase time period, as provided in an embodiment of the invention.

[0048] Figure 5This is a schematic diagram of the structure of a labor demand forecasting device provided in an embodiment of the present invention;

[0049] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0050] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] Chain retail enterprises, including large supermarkets and convenience stores, have numerous stores, a wide geographical reach, a large number of employees, and a high degree of complexity in store management. Given their large and complex organizational structure, achieving efficient management of store personnel is a pressing problem that chain retail enterprises need to solve, and staff scheduling is a crucial aspect of personnel management.

[0052] Currently, staff scheduling in supermarkets and stores is mainly done manually by store managers and department heads. Therefore, on the one hand, staff scheduling efficiency is low, and each scheduling takes 2 to 3 hours. On the other hand, the accuracy of scheduling depends on the experience of the store's scheduler, which can easily lead to redundancy or shortage of cashiers on certain dates or time periods. This results in low scheduling rationality. Redundancy of staff increases the ineffective working hours of cashiers, while shortage of staff affects the customer experience.

[0053] To facilitate understanding of this embodiment, a method for predicting labor demand disclosed in this invention will first be described in detail, such as... Figure 1 As shown, the method includes the following steps:

[0054] Step S102: Construct a time series decomposition model based on the multiple historical total sales data corresponding to the multiple historical time periods obtained in advance.

[0055] The aforementioned historical time period can be understood as a period of time before the current date, generally a certain day before the current date; multiple historical time periods can be understood as multiple periods of time before the current date, generally multiple days before the current date; each day before the current date, i.e. each historical time period, corresponds to a historical total sales data, and multiple days before the current date, i.e. multiple historical time periods, correspond to multiple historical total sales data.

[0056] In practice, historical sales data for the most recent month can be obtained based on the current date. If the month has 30 days, 30 historical total sales data points corresponding to each of the 30 days (30 historical time periods) can be obtained (with one historical total sales data point for each day). Then, a time series decomposition model can be constructed. Specifically, there is no limit to the length of the time span; it can be one year, six months, three months, one month, etc.

[0057] Specifically, to forecast offline sales (orders, SKUs, EAs) for physical stores, the Time Series Decomposition (STL) algorithm, also known as a time series decomposition model, can be used. This algorithm (model) can be understood as a sales forecasting model based on daily timeframes. It has low complexity, strong interpretability, and can fully consider the impact of promotions, holidays, etc., facilitating manual intervention. The Time Series Decomposition (STL) algorithm is mainly divided into two types:

[0058] One of them is the additive model decomposition algorithm:

[0059] For a time series {y(t), t=1,2…}, assuming it is an additive decomposition model, it can be written as:

[0060] y(t)=S(t)+T(t)+R(t)

[0061] Where y(t) is the original data, and S(t), T(t), and R(t) are the seasonal component, trend-cycle component, and residual component, respectively.

[0062] In actual implementation, assuming the current date is December 1, 2022, if we want to predict the offline sales for the next week (i.e., from December 1, 2022 to December 7, 2022), we can obtain the historical total sales data for each day of the previous month (November 1, 2022 to November 30, 2022), which is 30 days. The corresponding time series is {y(t), t=1,2…30}; where y(1) is the historical total sales data for November 1, 2022, y(2) is the historical total sales data for November 2, 2022, and so on, until y(30) is the historical total sales data for November 30, 2022.

[0063] Additive model decomposition steps:

[0064] Step 1: Long-term trend T(t) of the time series

[0065] Within the original time series, the average of several consecutive periods is successively calculated as the trend value of a certain period. By progressively shifting the averages in this way, a series of moving averages are obtained, forming an average time series.

[0066] If you want to predict the offline sales for the next week (December 1, 2022 to December 7, 2022), then the total historical sales data corresponding to the above consecutive periods, including 7 consecutive days, are used. Assuming that the original time series is still {y(t), t=1,2…30}, then the average of y(1) to y(7) can be used as the trend value of y(7); the average of y(2) to y(8) can be used as the trend value of y(8); and so on, until the average of y(24) to y(30) is obtained as the trend value of y(30);

[0067] The calculation method is as follows:

[0068] T(t)=(y t +y t-1 +…y t-N+1 ) / N

[0069] If the original time series is {y(t), t=1,2…30}, then N=7, and the average time series {T(t), t=7,8…30} is the trend value of {y(t), t=7,8…30}; where {y(t), t=7,8…30} is the sequence remaining after removing y(1), y(2), y(3), y(4), y(5) and y(6) from the original time series {y(t), t=1,2…30}.

[0070] In practice, it is necessary to start from November 1, 2022, i.e., y(1), and calculate the average of the historical total sales data for 7 consecutive days until November 30, 2022, i.e., y(30). Therefore, y(1), y(2), y(3), y(4), y(5) and y(6) do not have corresponding trend values.

[0071] Starting from y(7), there are corresponding trend values: the trend value T(7) for y(7), the trend value T(8) for y(8), and so on, until the trend value T(30) for y(30). Specifically, the trend values ​​from y(7) to y(30) can all be obtained using the above formula T(t)=(yt+yt-1+…yt-N+1) / N.

[0072] The following calculations are performed using the trend values ​​T(7), T(8), and T(30) corresponding to y(7), y(8), and y(30) as examples:

[0073] T(7)=(y 7+ y6+y5+y4+y 3+ y2+y1) / 7

[0074] T(8)=(y 8+ y 7+ y6+y5+y4+y 3+ y2) / 7

[0075] T(30)=(y 24+ y 2+ y 26 +y 27 +y 28 +y 29+ y 30 ) / 7

[0076] Based on the above, the correspondence between {y(t), t=1,2…30} and {T(t), t=7,8…30} is as follows:

[0077]

[0078]

[0079] Step 2: Calculate the time series D(t) = y(t) - T(t) after removing the trend component.

[0080] Because there are no T(1), T(2), T(3), T(4), T(5), T(6), there are also no D(1), D(2), D(3), D(4), D(5), D(6). Therefore, the time series without the trend component is {D(t), t=7,8…30}.

[0081] Step 3: Estimate the periodic component S(t)

[0082] Based on the above, the correspondence between {y(t), t=1,2…30} and {D(t), t=7,8…30} is as follows:

[0083]

[0084] In practice, the periodic components can be calculated by averaging the data within the same period. For example, given the original time series {y(t), t = 1, 2…30}, if we want to predict offline sales for the following week (December 1, 2022 to December 7, 2022) based on this original time series, we can average the data for Mondays, Tuesdays, Wednesdays, Thursdays, Mondays, Saturdays, and Sundays in all D(t). The periodic components are adjusted (by adding a bias) so that their sum is 0. The periodic components are then copied to the length of D(t), resulting in all the periodic components of D(t), denoted as S(t).

[0085] Specifically, the dates corresponding to D(7) to D(30) are November 7th to November 30th, 2022. By consulting the calendar, we know that there are four Mondays from November 7th to November 30th, 2022, with corresponding D(t) values ​​of D(7), D(14), D(21), and D(28). Therefore, the periodic component of Monday is S(1) = [D(7) + D(14) + D(21) + D(28)] / 4 There are four Tuesdays, with corresponding D(t) values ​​of D(8), D(15), D(22), and D(29). Therefore, the periodic component of Tuesday is S(2) = [D(8) + D(15) + D(22) + D(29)] / 4. There are four Wednesdays, with corresponding D(t) values ​​of D(9), D(16), D(23), and D(30). Therefore, the periodic component of Wednesday is S(3) = [D(9) + D(16) + D(29)] / 4. 3) + D(30)] / 3; There are three Thursdays, with corresponding D(t) values ​​of D(10), D(17), and D(24), respectively. Therefore, the periodic component of Thursday is S(4) = [D(10) + D(17) + D(24)] / 4; There are three Fridays, with corresponding D(t) values ​​of D(11), D(18), and D(25), respectively. Therefore, the periodic component of Friday is S(5) = [D(11) + D(18) + D(25)] / 3; There are three Saturdays, with corresponding D(t) values ​​of D(12), D(19), and D(26), respectively. Therefore, the periodic component of Saturday is S(6) = [D(12) + D(119) + D(26)] / 3; There are three Sundays, with corresponding D(t) values ​​of D(13), D(20), and D(27), respectively. Therefore, the periodic component of Sunday is S(7) = [D(13) + D(20) + D(27)] / 3.

[0086] Based on the above, in {S(t), t=1,2…7} and {D(t), t=7,8…30}, all Mondays have D(t) corresponding to S(1), all Tuesdays have D(t) corresponding to S(2), all Wednesdays have D(t) corresponding to S(3), all Thursdays have D(t) corresponding to S(4), all Fridays have D(t) corresponding to S(5), all Saturdays have D(t) corresponding to S(6), and all Sundays have D(t) corresponding to S(7).

[0087] Step 4: Residual component R(t) = y(t) - T(t) - S(t).

[0088] The residual component can be understood as the difference caused by the impact of factors such as holidays, promotions, and weather on the total historical sales data.

[0089] Based on the above, the correspondences between {y(t), t = 1, 2…30} and {D(t), t = 7, 8…30}, {S(t), t = 1, 2…7}, {R(t), t = 1, 2…30}, and the weekdays are as follows:

[0090]

[0091] Another time series decomposition algorithm (STL) is the multiplicative model decomposition algorithm. Similarly, a multiplicative model can be written as:

[0092] y(t)=S(t)×T(t)×R(t)

[0093] For a multiplicative model, we can take the logarithm (provided it is meaningful) to transform it into an additive model.

[0094] The ideas of multiplicative and additive models are very similar. The difference is that step two of the above additive model can be changed to D(t) = y(t) ÷ T(t), and step four can be changed to R(t) = y(t) ÷ (T(t) × S(t)).

[0095] The above method constructs a sales forecasting model based on daily time series decomposition. Compared to forecasting sales directly on an hourly basis, forecasting on a daily basis offers higher predictability and allows for the inclusion of various influencing factors, such as holidays, weather, and promotions.

[0096] In practice, by statistically analyzing the daily sales figures for a store across order, SKU, and EA dimensions from 2019 to 2020, we can observe that the SKU dimension exhibits the strongest statistical regularity (i.e., predictability), followed by the EA dimension, and then the order dimension. The order dimension can be understood as the number of orders; the SKU dimension as the number of product categories included in a single order; and the EA dimension as the number of products included in a single order. For example, if an order includes two bottles of water and one cake, then the corresponding SKU is 2, and the corresponding EA is 3.

[0097] Step S104: Based on the time series decomposition model, generate multiple predicted total sales data corresponding to multiple predicted time periods.

[0098] As can be seen from step S102, there are two types of time series decomposition models. Assuming y(t) = S(t) + T(t) + R(t), and the current date is December 1, 2022, if we want to predict the total sales data for the next week (December 1, 2022 to December 7, 2022), which is seven days (seven prediction time periods), we can predict according to y(t) = S(t) + T(t) + R(t). T(t) can be the mean of the week closest to the current date, which is T(30). S(t) can be the corresponding periodic component based on the weekdays from December 1, 2022 to December 7, 2022. R(t) can be further determined based on the weather from December 1, 2022 to December 7, 2022, as well as whether there are holidays, promotional activities, etc.

[0099] Specifically, assuming that the seven predicted total sales data corresponding to December 1, 2022 to December 7, 2022 (Thursday to Wednesday) are y(31), y(32), y(33), y(34), y(35), y(36), and y(37), if we do not consider the influence of weather, holidays, and promotional activities, then y(31) = S(4) + T(30) + R(30), y(32) = S(5) + T(30) + R(30), y(33) = S(6) + T(30) + R(30), y(34) = S(7) + T(30) + R(30), y(35) = S(1) + T(30) + R(30), y(36) = S(2) + T(30) + R(30), and y(37) = S(3) + T(30) + R(30).

[0100] If we consider the impact of weather, holidays, and promotional activities, taking the weather as an example, assuming that there are rainy days from December 1 to December 7, 2022, we can first obtain the R(t) corresponding to all rainy days from November 1 to November 30, 2022, and then calculate the average value. This average value is used as the R(t) when calculating y(31), y(32), y(33), y(34), y(35), y(36), and y(37).

[0101] Step S106: For each predicted total sales data, based on the predicted total sales data and multiple pre-acquired average ratios, divide the predicted total sales data into multiple predicted sub-sales data corresponding to each predicted sub-time period.

[0102] Each of the above-mentioned forecast sub-time periods can be understood as one hour. Each forecast time period can include multiple forecast sub-time periods, equivalent to daily working hours. For example, if the prescribed working hours are from 8:00 AM to 10:00 PM, with no lunch break, then there are 14 forecast sub-time periods per day. The number of forecast sub-time periods included in each forecast time period (per day) is usually the same. Each of the above-mentioned average proportions can be understood as the percentage of historical sub-sales data in the corresponding historical sub-time period relative to the total historical sales data within the historical time period.

[0103] In practice, to improve the accuracy of forecasts and thus arrange appropriate working hours and reasonable shift scheduling, offline sales (orders, SKUs, or EAs) of stores can be predicted on an hourly basis. The sales of a day (total predicted sales data) can be divided into sales of multiple time periods on an hourly basis (multiple predicted sub-sales data) according to a certain proportion (multiple average proportions), and a forecast result for a week can be given at once.

[0104] Assuming the predicted total sales data for December 1, 2022 is y(31) = 1000, and December 1, 2022 includes 14 predicted sub-time periods, the average proportions of the 14 predicted sub-time periods obtained in advance are as follows: 8:00-9:00 is 4.49%, 9:00-10:00 is 8.24%, 10:00-11:00 is 9.83%, 11:00-12:00 is 9.08%, 12:00-1... The percentages for different time periods are as follows: 3:00 - 6.62%, 13:00-14:00 - 5.52%, 14:00-15:00 - 6.47%, 15:00-16:00 - 7.87%, 16:00-17:00 - 8.60%, 17:00-18:00 - 8.51%, 18:00-19:00 - 7.56%, 19:00-20:00 - 7.10%, and 20:00-21:00 - 6.53%. The 21:00-22:00 forecast is 3.58%. Therefore, the predicted sales figures for the 14 predicted sub-time periods are as follows: 8:00-9:00: 44.9 (1000 × 4.49% = 44.9); 9:00-10:00: 82.4 (1000 × 8.24% = 82.4); 10:00-11:00: 98.3; 11:00-12:00: 90.8; 12:00-13:00: 66. 2%, 13:00-14:00 is 55.2, 14:00-15:00 is 64.7, 15:00-16:00 is 78.7, 16:00-17:00 is 86.0, 17:00-18:00 is 85.1, 18:00-19:00 is 75.6, 19:00-20:00 is 71.0, 20:00-21:00 is 65.3, 21:00-22:00 is 35.8.

[0105] In practice, by obtaining the average sales percentage (percentage of daily sales during that period) for a store from June 2019 to March 2021 across order, SKU, and EA dimensions, a clear pattern emerges: the daily trends are roughly consistent, first rising, then falling, then rising again, and then falling again. The two peak periods are 10:00-11:00 and 16:00-17:00, respectively. Furthermore, the average sales percentage changes little within the same time period, such as 8:00-9:00 each day, and can be approximated as the same. Of course, the average sales percentage may vary during specific time periods, so factors such as weather, promotions, and holidays should be considered, and adjustments should be made based on the actual situation and the average sales percentage.

[0106] Step S108: Based on the sales data of each predicted sub-segment, generate the labor demand information corresponding to each predicted sub-time period.

[0107] In practice, this can be understood as intelligent scheduling based on sales volume during a time period and relevant scheduling rules. For example, if the predicted sales volume for a certain sub-time period (16:00-17:00) is 86 orders, and the scheduling rules stipulate that each cashier can handle a maximum of 45 orders per hour (which can also be understood as 1 work hour), then two cashiers can be arranged for this predicted sub-time period (8:00-9:00), which means the corresponding labor demand information (number of cashiers or scheduling work hours) is 2.

[0108] Human resource efficiency is a very important core indicator for retail enterprises. Reasonable staffing based on the store's operating conditions can improve work efficiency, reduce labor costs, and enhance the ability to survive in competition. Reasonable staff scheduling is an effective way to improve human resource efficiency.

[0109] Standard daily productivity = Daily sales / (Total actual attendance and clock-in hours of employees / Standard daily working hours) / 1000

[0110] Standard monthly employee productivity = Previous month's sales revenue / (Employee's total working hours in the previous month / Standard monthly working hours) / 1000

[0111] Store productivity is directly related to store sales and store shift hours. Increasing sales and optimizing store shifts can improve store productivity.

[0112] like Figure 2 The diagram illustrates the correlation between store labor efficiency, store size, sales revenue, and shift work hours. A comparative analysis of labor efficiency across stores of similar size reveals a generally positive correlation between shift work hours and sales revenue for stores of the same size. Based on this observation, a rationality analysis of store shift work can be conducted.

[0113] (1) High store sales volume, short shift hours, and high labor efficiency

[0114] While this approach offers high efficiency, it's important to consider whether employees are under excessive work pressure. Excessive work pressure can negatively impact customer experience and potentially lead to high employee turnover.

[0115] (2) Low store sales, high shift work hours, and low labor efficiency

[0116] This situation indicates that the workload of employees is relatively small, so the number of shifts can be reduced appropriately to lower labor costs and make the store more competitive.

[0117] The above-described embodiments of this application can accurately predict future sales, thereby enabling reasonable staff scheduling. On the one hand, this can save time spent on manual scheduling, and on the other hand, it can alleviate the situation where unreasonable scheduling leads to redundant staff, resulting in increased ineffective working hours, or insufficient staff, which affects customer experience and increases employee stress.

[0118] The aforementioned method for forecasting labor demand includes: constructing a time series decomposition model based on multiple historical total sales data corresponding to multiple pre-acquired historical time periods; generating multiple predicted total sales data corresponding to multiple forecast time periods based on the time series decomposition model; dividing the predicted total sales data into multiple predicted sub-sales data corresponding to each forecast sub-time period based on the predicted total sales data and multiple pre-acquired average proportions; and generating labor demand information corresponding to each forecast sub-time period based on each predicted sub-sales data. This method, by dividing the predicted total sales data of multiple forecast time periods generated by the time series decomposition model according to their respective average proportions, can predict future sales, thereby enabling reasonable personnel scheduling, improving the rationality of scheduling, and avoiding personnel redundancy or shortage.

[0119] This invention also provides another method for predicting labor demand, which is implemented based on the method described in the above embodiments; such as Figure 3 As shown, the method includes the following steps:

[0120] Step S202: Construct a time series decomposition model based on the multiple historical total sales data corresponding to the multiple historical time periods obtained in advance.

[0121] In practice, each historical time period includes multiple historical sub-time periods; for each historical time period, the total historical sales data corresponding to that historical time period is the sum of the historical sub-sales data corresponding to each historical sub-time period of that historical time period.

[0122] Specifically, each historical time period generally includes 14 historical sub-time periods. Assuming one of the historical time periods is November 23, 2022, the 14 historical sub-time periods are (8:00-22:00). From 8:00 to 22:00, each hour corresponds to a historical sub-time period. The sales volume of each historical sub-time period (e.g., 8:00-9:00) is the historical sub-sales data corresponding to that historical sub-time period. Then, the sum of the historical sub-sales data corresponding to the 14 historical sub-time periods from 8:00 to 22:00 is the historical total sales data corresponding to that historical time period.

[0123] Step S204: Based on the time series decomposition model, generate multiple predicted total sales data corresponding to multiple predicted time periods.

[0124] Step S206: For each predicted total sales data, based on the predicted total sales data and multiple pre-acquired average ratios, divide the predicted total sales data into multiple predicted sub-sales data corresponding to each predicted sub-time period.

[0125] Specifically, each average proportion can be calculated through steps five through seven:

[0126] Step 5: Obtain the historical sub-sales data corresponding to each historical sub-time period.

[0127] In actual implementation, it is assumed that multiple historical time periods are from November 1, 2022 to November 30, 2022, that is, 30 historical time periods. Each of these 30 historical time periods can include 14 historical sub-time periods (that is, 8:00-22:00 in a day); then the historical sub-sales data corresponding to the 14 historical sub-time periods included in each of the 30 historical time periods are obtained.

[0128] Step 6: Based on each historical sub-sales data and each historical total sales data, obtain the proportion of historical sub-sales data corresponding to each historical sub-time period.

[0129] In actual implementation, based on the current date, the 30 most recent historical total sales data points and the 420 historical sub-sales data points (time period sales data) corresponding to the 30 historical total sales data points can be obtained. Then, based on these 420 historical sub-sales data points and the 30 historical total sales data points (each historical total sales data point corresponds to 14 historical sub-sales data points), the proportion of each historical sub-sales data point to its corresponding historical total sales data point can be calculated (equivalent to the proportion of historical sub-sales data points corresponding to each historical sub-time period), thus obtaining the proportion of the 420 historical sub-sales data points.

[0130] The aforementioned 420 historical sub-sales data ratios include 30 ratios corresponding to 8:00-9:00 (8:00-9:00 every day of the 30 days from November 1, 2022 to November 30, 2022), 30 ratios corresponding to 9:00-10:00, 30 ratios corresponding to 10:00-11:00, 30 ratios corresponding to 11:00-12:00, 30 ratios corresponding to 12:00-13:00, and 30 ratios corresponding to 13:00-14:00. The percentages of historical sub-sales data corresponding to 14:00-15:00, 15:00-16:00, 16:00-17:00, 17:00-18:00, 18:00-19:00, 19:00-20:00, 20:00-21:00, and 21:00-22:00.

[0131] Step 7: Based on the proportion of each historical sub-sales data, determine the average proportion corresponding to each historical sub-time period.

[0132] In practice, the average of the historical sub-sales data ratios corresponding to 30 periods from 8:00 to 9:00 can be calculated as the average ratio for the historical sub-time period of 8:00 to 9:00; the average of the historical sub-sales data ratios corresponding to 30 periods from 9:00 to 10:00 can be calculated as the average ratio for the historical sub-time period of 9:00 to 10:00; the average of the historical sub-sales data ratios corresponding to 30 periods from 10:00 to 11:00 can be calculated as the average ratio for the historical sub-time period of 10:00 to 11:00; and the average of the historical sub-sales data ratios corresponding to 30 periods from 11:00 to 12:00 can be calculated as the average ratio for the historical sub-time period of 10:00 to 11:00. The average of the 30 historical sub-time periods from 11:00 to 12:00 was used as the average proportion for the 11:00-12:00 period; the average of the 30 historical sub-time periods from 12:00 to 13:00 was used as the average proportion for the 12:00-13:00 period; the average of the 30 historical sub-time periods from 13:00 to 14:00 was used as the average proportion for the 13:00-14:00 period; and the average of the 30 historical sub-time periods from 14:00 to 15:00 was used as the average proportion for the 14:00-15:00 period. For example, calculate the average of the historical sub-sales data ratios corresponding to 30 periods from 15:00 to 16:00 as the average ratio for the historical sub-time period of 15:00 to 16:00; calculate the average of the historical sub-sales data ratios corresponding to 30 periods from 16:00 to 17:00 as the average ratio for the historical sub-time period of 16:00 to 17:00; calculate the average of the historical sub-sales data ratios corresponding to 30 periods from 17:00 to 18:00 as the average ratio for the historical sub-time period of 17:00 to 18:00; calculate the average of the historical sub-sales data ratios corresponding to 30 periods from 18:00 to 19:00. The average value is used as the average proportion corresponding to the historical sub-time period of 18:00-19:00; the average proportion of the historical sub-sales data corresponding to 30 times 19:00-20:00 is used as the average proportion corresponding to the historical sub-time period of 19:00-20:00; the average proportion of the historical sub-sales data corresponding to 30 times 20:00-21:00 is used as the average proportion corresponding to the historical sub-time period of 20:00-21:00; the average proportion of the historical sub-sales data corresponding to 30 times 21:00-22:00 is used as the average proportion corresponding to the historical sub-time period of 21:00-22:00.

[0133] Assuming we obtain 7 predicted total sales data points, for each predicted total sales data point, we can multiply it by the average proportion corresponding to the aforementioned 14 historical sub-time periods to obtain predicted sub-sales data for each of the 14 predicted sub-time periods. The 14 historical sub-time periods are the same as the 14 predicted sub-time periods: 8:00-9:00, 9:00-10:00, 10:00-11:00, 11:00-12:00, 12:00-13:00, 13:00-14:00, 14:00-15:00, 15:00-16:00, 16:00-17:00, 17:00-18:00, 18:00-19:00, 19:00-20:00, 20:00-21:00, and 21:00-22:00.

[0134] Step S208: Obtain the target historical smart purchase sales ratio corresponding to each historical sub-time period.

[0135] In practice, cashiers are an important part of supermarket and retail store operations. Demand fluctuates greatly, ranging from major holidays throughout the year to peak and off-peak periods within a single day. Furthermore, with the introduction of various smart shopping devices, including self-service checkout, independent checkout, and smart shopping carts, scheduling needs to balance smart shopping and manual checkout, making the scheduling of cashiers even more complex.

[0136] Specifically, each historical sub-sales data point can include smart checkout sales data and manual checkout sales data. The proportion of each smart checkout sales data point to its corresponding historical sub-sales data is called the historical smart checkout sales proportion. Smart checkout can be understood as the core of multi-point checkout products, providing multiple checkout methods to meet the checkout needs of various offline scenarios, greatly improving checkout efficiency and enhancing user experience. This includes DPOS, self-service checkout, independent checkout, and smart shopping carts. Manual checkout can be understood as the manual checkout counters in supermarkets and convenience stores.

[0137] See Figure 4 The diagram shows a sales volume over a specific time period (historical sub-sales data) and the sales ratio of the smart purchase time period (historical smart purchase sales ratio). Specifically... Figure 4 This data represents the sales volume and the percentage of sales during the smart shopping period for a certain store during the week from March 12th to March 18th, 2020. Figure 4 It can be seen that there are two peaks in the sales ratio of smart shopping during different time periods: one is from 11:00 to 13:00, and the other is after 16:00; the peak of the sales ratio of smart shopping lags behind the peak of the sales during different time periods.

[0138] Specifically, step S208 can be achieved through the following steps eight to ten:

[0139] Step 8: Obtain the historical smart purchase and sales data corresponding to each historical sub-time period.

[0140] Assuming 30 historical time periods are obtained, each historical time period can have 14 historical sub-time periods. The historical sub-sales data corresponding to each historical sub-time period is the sum of the smart purchase sales data (equivalent to historical smart purchase sales data) and the manual bank sales data corresponding to that historical sub-time period.

[0141] In actual implementation, the historical smart purchase and sales data corresponding to the 14 historical sub-time periods included in each of the 30 historical time periods can be obtained, resulting in 420 historical smart purchase and sales data.

[0142] Step 9: Based on each historical smart purchase sales data and each historical sub-sales data, obtain the historical smart purchase sales ratio corresponding to each historical sub-time period.

[0143] In actual implementation, the proportion of 420 historical smart purchase sales data to their corresponding historical sub-sales data can be calculated (equivalent to the historical smart purchase sales proportion corresponding to each historical sub-time period), thus obtaining the 420 historical smart purchase sales proportions.

[0144] Step 10: Based on the historical smart purchase sales ratio, obtain the target historical smart purchase sales ratio corresponding to each historical sub-time period.

[0145] In practice, the aforementioned 420 historical intelligent purchase and sales ratios include: 30 historical intelligent purchase and sales ratios corresponding to 8:00-9:00 (each day from 8:00-9:00 during the 30 days from November 1st to November 30th, 2022); 30 historical intelligent purchase and sales ratios corresponding to 9:00-10:00; 30 historical intelligent purchase and sales ratios corresponding to 10:00-11:00; 30 historical intelligent purchase and sales ratios corresponding to 11:00-12:00; 30 historical intelligent purchase and sales ratios corresponding to 12:00-13:00; and 30 historical intelligent purchase and sales ratios corresponding to 13:00-14:00. Volume ratio, historical smart purchase sales ratio for 30 periods from 14:00 to 15:00, historical smart purchase sales ratio for 30 periods from 15:00 to 16:00, historical smart purchase sales ratio for 30 periods from 16:00 to 17:00, historical smart purchase sales ratio for 30 periods from 17:00 to 18:00, historical smart purchase sales ratio for 30 periods from 18:00 to 19:00, historical smart purchase sales ratio for 30 periods from 19:00 to 20:00, historical smart purchase sales ratio for 30 periods from 20:00 to 21:00, and historical smart purchase sales ratio for 30 periods from 21:00 to 22:00.

[0146] Specifically, the target historical smart purchase sales ratio can be selected from 30 historical smart purchase sales ratios corresponding to 8:00-9:00 (equivalent to the highest smart purchase ratio during the same period) as the 8:00-9:00 sub-time period; the target historical smart purchase sales ratio can be selected from 30 historical smart purchase sales ratios corresponding to 9:00-10:00; the target historical smart purchase sales ratio can be selected from 30 historical smart purchase sales ratios corresponding to 10:00-11:00; the target historical smart purchase sales ratio can be selected from 30 historical smart purchase sales ratios corresponding to 11:00-12:00. The highest value among the sales ratios is selected as the target historical smart purchase sales ratio for the historical sub-time period of 11:00-12:00; the highest value among the 30 historical smart purchase sales ratios corresponding to 12:00-13:00 is selected as the target historical smart purchase sales ratio for the historical sub-time period of 12:00-13:00; the highest value among the 30 historical smart purchase sales ratios corresponding to 13:00-14:00 is selected as the target historical smart purchase sales ratio for the historical sub-time period of 13:00-14:00; the highest value among the 30 historical smart purchase sales ratios corresponding to 14:00-15:00 is selected as the target historical smart purchase sales ratio for the historical sub-time period of 14:00-15:00. The target historical smart purchase sales ratio is calculated as follows: The highest value is selected from 30 historical smart purchase sales ratios corresponding to 15:00-16:00 as the target historical smart purchase sales ratio for the 15:00-16:00 sub-period; the highest value is selected from 30 historical smart purchase sales ratios corresponding to 16:00-17:00 as the target historical smart purchase sales ratio for the 16:00-17:00 sub-period; the highest value is selected from 30 historical smart purchase sales ratios corresponding to 17:00-18:00 as the target historical smart purchase sales ratio for the 17:00-18:00 sub-period; the highest value is selected from 30 historical smart purchase sales ratios corresponding to 18:00-19 ... The maximum value is used as the target historical smart purchase sales ratio for the historical sub-time period of 18:00-19:00; the maximum value is selected from 30 historical smart purchase sales ratios corresponding to 19:00-20:00 as the target historical smart purchase sales ratio for the historical sub-time period of 19:00-20:00; the maximum value is selected from 30 historical smart purchase sales ratios corresponding to 20:00-21:00 as the target historical smart purchase sales ratio for the historical sub-time period of 20:00-21:00; the maximum value is selected from 30 historical smart purchase sales ratios corresponding to 21:00-22:00 as the target historical smart purchase sales ratio for the historical sub-time period of 21:00-22:00.

[0147] Step S210: Based on the sales data of each predicted sub-sales segment and the historical intelligent purchase sales ratio of each target, determine the manual bank sales data corresponding to each predicted sub-time period.

[0148] Specifically, step S210 can be achieved through the following steps eleven to twelve:

[0149] Step 11: Based on the sales data of each predicted sub-segment and the historical smart purchase sales ratio of each target, determine the predicted smart purchase sales data corresponding to each predicted sub-period.

[0150] In actual implementation, taking a certain prediction period (e.g., December 1, 2022) as an example, this prediction period includes 14 prediction sub-periods (the same as the 14 historical sub-periods). Assuming that we obtain the 14 prediction sub-sales data corresponding to the 14 prediction sub-periods, and that these 14 prediction sub-sales data correspond to the aforementioned 14 target historical smart purchase sales ratios, we can multiply these 14 prediction sub-sales data by their corresponding target historical smart purchase sales ratios to obtain the prediction smart purchase sales data corresponding to the 14 prediction sub-periods.

[0151] Step 12: Calculate the difference between each predicted sub-sales data and each predicted intelligent purchase sales data to obtain the manual bank sales data corresponding to each predicted sub-time period.

[0152] In actual implementation, taking a certain prediction period (such as December 1, 2022) as an example, by calculating the difference between the above 14 predicted sub-sales data and their corresponding predicted smart purchase sales data, we can obtain the manual bank sales data corresponding to the 14 predicted sub-sales periods.

[0153] Specifically, the workload of manual checkout counters (manual checkout sales data for each predicted sub-period) equals the total sales volume for the period (predicted sub-period sales data for each predicted sub-period) minus the sales volume of smart shopping (predicted smart shopping sales data for each predicted sub-period). For example, if the predicted total sales volume for the period 8:00-9:00 on December 1, 2022 is 100 and the smart shopping sales volume is 20 (i.e., the target historical smart shopping sales volume ratio for the historical sub-period of 8:00-9:00 is 20%), then the workload of manual checkout counters is 80 (100-20).

[0154] Step S212: Based on the sales data of each ICBC branch, generate the labor demand information corresponding to each predicted sub-time period.

[0155] In practice, if the workload of the manual cashier during the 8:00-9:00 shift on December 1, 2022 is 80, and each cashier can handle a maximum of 45 orders per work hour, then at least two cashiers are needed to complete the task. Therefore, two cashiers can be scheduled during the 8:00-9:00 shift on December 1, 2022.

[0156] Specifically, the above embodiments can be understood as a labor demand forecasting method based on maximizing the efficiency of smart purchasing. The goal of maximizing the efficiency of smart purchasing is to maximize the sales ratio of smart purchasing to offline orders in each time period of the day. The highest proportion of smart purchasing in the same period in history (e.g., within the most recent month) can be used as the efficiency for that period for calculation. This is equivalent to overestimating the sales ratio of smart purchasing to a certain extent and underestimating the sales volume of manual cashiers, thereby reducing labor demand, reducing labor costs, and making work schedules more reasonable.

[0157] The aforementioned method for forecasting labor demand combines historical total sales data, historical sub-sales data, historical sub-sales data ratios, historical smart purchase sales data, historical smart purchase sales ratios, and factors such as holidays, promotions, and weather to predict future store sales (manual cashier sales data) in an automated and intelligent manner. This allows for the estimation of future labor demand, providing data support for intelligent scheduling of store staff. It balances smart purchase and manual cashier operations, improving the rationality of cashier scheduling. The method described in this paper is also applicable to forecasting labor demand for other positions.

[0158] Step S214: Based on the pre-set evaluation indicators, evaluate the rationality of the employment demand information corresponding to each predicted sub-time period.

[0159] The aforementioned evaluation indicators include the target average order volume per unit of working hours and the target coefficient of variation. In practice, to evaluate the rationality of the current store's scheduling, we can first define evaluation indicators for scheduling rationality. By using quantitative indicators, we can objectively, reasonably, and uniformly evaluate the scheduling situation, identify problems in the current store's scheduling, and make targeted optimizations. Specifically, we can first calculate the average order volume per unit of working hours for the store within a certain time range (week, month, etc.), and the coefficient of variation of the average order volume per unit of working hours. Then, we compare these two evaluation indicators with the target average order volume per unit of working hours and the target coefficient of variation to measure the rationality of the store's scheduling. The coefficient of variation represents the degree of fluctuation (stability).

[0160] Specifically, step S214 can be achieved through steps thirteen to nineteen:

[0161] Step 13: Obtain the labor demand information corresponding to each predicted sub-time period and the actual labor sales data corresponding to each predicted time period.

[0162] In practice, scheduling can be generated based on the labor demand information corresponding to each predicted sub-time period, and cashiers can then be assigned tasks according to this scheduling data. Although the above embodiment can predict the sales data of manual cash registers for each predicted sub-time period, in actual store operation, due to various unforeseen factors, the actual sales data of manual cash registers may differ significantly from the predicted data, resulting in unreasonable scheduling. Therefore, after the actual operation of the store is completed, the rationality of the scheduling data needs to be evaluated.

[0163] Step Fourteen: Based on the labor demand information corresponding to each predicted sub-time period, obtain the total labor demand information corresponding to each predicted time period.

[0164] In practice, assuming that we obtain the labor demand information (shift work hours) corresponding to the 14 predicted sub-time periods in each of the seven predicted time periods (December 1, 2022 to December 7, 2022), then for each predicted time period, we can calculate the sum of the labor demand information corresponding to the 14 predicted sub-time periods to obtain the total labor demand information (daily shift work hours) corresponding to the seven predicted time periods.

[0165] Step 15: Based on the actual sales data of each bank counter and the total labor demand information, determine the order volume per unit of working hours for each forecast period.

[0166] In practice, for each predicted time period, the ratio of the corresponding actual manual bank counter sales data (daily order volume) to the corresponding total labor demand information (daily shift hours) can be calculated to obtain the order volume processed per unit of working hours, as shown below:

[0167]

[0168] Assuming the actual daily order volume (dayly sales volume) of bank tellers for the seven forecast periods from December 1st to December 7th, 2022 is 700, 800, 1000, 900, 1200, 1400, and 1600 respectively; and the total labor demand (daily shift hours) for the seven forecast periods from December 1st to December 7th, 2022 is 35, 20, 40, 30, 25, 40, and 50 respectively; then the order volume per unit hour for the seven forecast periods from December 1st to December 7th, 2022 is 20, 40, 25, 30, 48, 35, and 32 respectively.

[0169] Step 16: Determine the average number of orders processed per unit of time based on the number of orders processed per unit of time for each forecast period.

[0170] Based on the order volume per unit working hour corresponding to the above 7 predicted time periods, the average order volume per unit working hour can be calculated to be 33, that is, (20+40+25+30+48+35+32)÷7=33. When calculating, the integer should be retained according to the rounding principle.

[0171] Step 17: Determine the standard deviation of the order volume processed per unit of time based on the average order volume processed per unit of time and the order volume processed per unit of time for each forecast period.

[0172] In practice, we can first calculate the square of the absolute value of the number of orders processed per unit of time and the average number of orders processed per unit of time, then calculate the sum of the squares of multiple absolute values, and finally take the square root of the sum of squares to obtain the standard deviation of the number of orders processed per unit of time, as shown below:

[0173]

[0174] Where i is the i-th prediction time period; a is the average number of orders processed per unit of working hours; fi is the number of orders processed per unit of working hours corresponding to the i-th prediction time period; and b is the standard deviation of the number of orders processed per unit of working hours.

[0175] Therefore, when the order volume per unit of working hours for the seven forecast periods from December 1st to December 7th, 2022 are 20, 40, 25, 30, 48, 35, and 32 respectively, and the average order volume per unit of working hours is 33, the standard deviation of the order volume per unit of working hours can be calculated to be 23, as shown below:

[0176]

[0177] Step 18: Calculate the ratio of the standard deviation of the number of orders processed per unit of working hours to the mean of the number of orders processed per unit of working hours to obtain the coefficient of variation.

[0178] In actual implementation, the formula is as follows:

[0179]

[0180] Therefore, when the standard deviation of the number of orders processed per unit of working hours is 23 and the mean of the number of orders processed per unit of working hours is 33, the coefficient of variation is 0.7. When calculating, according to the rounding principle, one decimal place should be retained.

[0181] Step 19: If the average number of orders processed per unit of working hours is greater than the target average number of orders processed per unit of working hours and the coefficient of variation is less than the target coefficient of variation, then the labor demand information is deemed reasonable.

[0182] In practice, the target average order volume per unit of working hours and the target coefficient of variation can be preset. If the average order volume per unit of working hours is less than the target average order volume per unit of working hours, or the coefficient of variation is greater than the target coefficient of variation, or if the average order volume per unit of working hours is less than the target average order volume per unit of working hours and the coefficient of variation is greater than the target coefficient of variation, it can be determined that the labor demand information is unreasonable, thus determining that the scheduling data is unreasonable. If the average order volume per unit of working hours is greater than the target average order volume per unit of working hours and the coefficient of variation is less than the target coefficient of variation, it can be determined that the labor demand information is reasonable, thus determining that the scheduling data is reasonable.

[0183] When the coefficient of variation (COP) is less than the target COP, it indicates that the daily scheduling of employees matches the trend of order volume changes; that is, more shifts are scheduled when there are more orders and fewer shifts are scheduled when there are fewer orders. When the COP is greater than the target COP, it indicates that the daily scheduling of employees matches the trend of order volume changes. When the average number of orders processed per unit of time is greater than the target average number of orders processed per unit of time, it indicates that the workload is saturated. Conversely, when the average number of orders processed per unit of time is less than the target average number of orders processed per unit of time, it indicates that the scheduled working hours are too many and the workload of employees is not saturated.

[0184] Therefore, when the shift schedule does not match the order volume, the shift schedule can be optimized to improve its rationality and thus increase the store's labor efficiency. A reasonable shift schedule depends on accurate order volume forecasting. When the order volume per unit of working hours is low, the reasons can be analyzed to see if the number of people (shift hours) can be appropriately reduced to save labor costs.

[0185] The aforementioned labor demand forecasting method provides quantitative indicators for evaluating the rationality of scheduling. Stores can use these indicators to monitor the rationality of scheduling and identify scheduling problems. At the same time, the method runs automatically, enabling sales forecasting and automatic scheduling for large-scale stores. Furthermore, the method can continuously optimize the algorithm based on business feedback.

[0186] This invention also provides a labor demand forecasting device, such as... Figure 5 As shown, the device includes: a construction module 50, used to construct a time series decomposition model based on multiple historical total sales data corresponding to multiple historical time periods obtained in advance; a first generation module 51, used to generate multiple predicted total sales data corresponding to multiple predicted time periods based on the time series decomposition model; a division module 52, used to divide the predicted total sales data into multiple predicted sub-sales data corresponding to each predicted sub-time period based on the predicted total sales data and multiple average ratios obtained in advance; and a second generation module 53, used to generate labor demand information corresponding to each predicted sub-time period based on each predicted sub-sales data.

[0187] The aforementioned labor demand forecasting device includes: constructing a time series decomposition model based on multiple historical total sales data corresponding to multiple historical time periods obtained in advance; generating multiple predicted total sales data corresponding to multiple prediction time periods based on the time series decomposition model; for each predicted total sales data, dividing the predicted total sales data into multiple predicted sub-sales data corresponding to each prediction sub-time period based on the predicted total sales data and multiple pre-obtained average ratios; and generating labor demand information corresponding to each prediction sub-sales data. This device, by dividing the predicted total sales data of multiple prediction time periods generated by the time series decomposition model according to their respective average ratios, can predict future sales, thereby enabling reasonable personnel scheduling, improving the rationality of scheduling, and avoiding personnel redundancy or shortage.

[0188] Furthermore, each historical time period includes multiple historical sub-time periods; for each historical time period, the historical total sales data corresponding to that historical time period is the sum of the historical sub-sales data corresponding to each historical sub-time period of that historical time period; the device also includes an acquisition module, which is used to acquire the historical sub-sales data corresponding to each historical sub-time period; based on each historical sub-sales data and each historical total sales data, the proportion of historical sub-sales data corresponding to each historical sub-time period is obtained; based on the proportion of each historical sub-sales data, the average proportion corresponding to each historical sub-time period is determined.

[0189] Furthermore, the second generation module is also used to: obtain the target historical intelligent purchase and sales ratio corresponding to each historical sub-time period; determine the manual bank sales data corresponding to each predicted sub-time period based on each predicted sub-time period and each target historical intelligent purchase and sales ratio; and generate the labor demand information corresponding to each predicted sub-time period based on each manual bank sales data.

[0190] Furthermore, the second generation module is also used to: obtain the historical smart purchase sales data corresponding to each historical sub-time period; obtain the historical smart purchase sales ratio corresponding to each historical sub-time period based on each historical smart purchase sales data and each historical sub-sales data; and obtain the target historical smart purchase sales ratio corresponding to each historical sub-time period based on each historical smart purchase sales ratio.

[0191] Furthermore, the second generation module is also used to: determine the predicted intelligent purchase sales data corresponding to each predicted sub-period based on each predicted sub-sales data and each target historical intelligent purchase sales ratio; calculate the difference between each predicted sub-sales data and each predicted intelligent purchase sales data to obtain the manual bank sales data corresponding to each predicted sub-period.

[0192] Furthermore, the device also includes an evaluation module, which is used to evaluate the rationality of the labor demand information corresponding to each predicted sub-time period based on pre-set evaluation indicators.

[0193] Furthermore, the evaluation indicators include the target average order volume per unit of working hours and the target coefficient of variation. The evaluation module is also used to: obtain the labor demand information corresponding to each prediction sub-period and the actual manual bank counter sales data corresponding to each prediction period; based on the labor demand information corresponding to each prediction sub-period, obtain the total labor demand information corresponding to each prediction period; based on each actual manual bank counter sales data and each total labor demand information, determine the order volume per unit of working hours corresponding to each prediction period; based on the order volume per unit of working hours corresponding to each prediction period, determine the average order volume per unit of working hours; based on the average order volume per unit of working hours and the order volume per unit of working hours corresponding to each prediction period, determine the standard deviation of the order volume per unit of working hours; calculate the ratio of the standard deviation of the order volume per unit of working hours to the average order volume per unit of working hours to obtain the coefficient of variation; if the average order volume per unit of working hours is greater than the target average order volume per unit of working hours and the coefficient of variation is less than the target coefficient of variation, the labor demand information is deemed reasonable.

[0194] The labor demand forecasting device provided in this embodiment of the invention has the same implementation principle and technical effect as the aforementioned labor demand forecasting method embodiment. For the labor demand forecasting device embodiment, please refer to the corresponding content in the aforementioned labor demand forecasting method embodiment.

[0195] This invention also provides an electronic device, see [link to relevant documentation]. Figure 6 As shown, the electronic device includes a processor 130 and a memory 131. The memory 131 stores machine-executable instructions that can be executed by the processor 130. The processor 130 executes the machine-executable instructions to implement the above-mentioned labor demand forecasting method.

[0196] Furthermore, Figure 6 The electronic device shown also includes a bus 132 and a communication interface 133, with the processor 130, the communication interface 133 and the memory 131 connected via the bus 132.

[0197] The memory 131 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 133 (which can be wired or wireless), such as the Internet, wide area network, local area network, or metropolitan area network. The bus 132 may be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 6 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0198] Processor 130 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 130 or by instructions in software form. Processor 130 may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 131, and processor 130 reads the information in memory 131 and, in conjunction with its hardware, completes the steps of the method described in the foregoing embodiments.

[0199] This invention also provides a computer-readable storage medium storing computer-executable instructions. When these computer-executable instructions are invoked and executed by a processor, they cause the processor to implement the aforementioned labor demand prediction method. For specific implementation details, please refer to the method embodiments, which will not be repeated here.

[0200] The labor demand forecasting method, apparatus, and electronic device provided in the embodiments of the present invention include a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the preceding method embodiments. For specific implementation, please refer to the method embodiments, which will not be repeated here.

[0201] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0202] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for forecasting labor demand, characterized in that, The method includes: A time series decomposition model is constructed based on multiple historical total sales data corresponding to multiple historical time periods obtained in advance. Based on the time series decomposition model, multiple predicted total sales data corresponding to multiple predicted time periods are generated. For each of the predicted total sales data, based on the predicted total sales data and multiple pre-acquired average ratios, the predicted total sales data is divided into multiple predicted sub-sales data corresponding to each predicted sub-time period. Based on the sales data of each predicted sub-segment, generate the labor demand information corresponding to each predicted sub-time period; Each historical time period includes multiple historical sub-time periods; for each historical time period, the total historical sales data corresponding to that historical time period is the sum of the historical sub-sales data corresponding to each historical sub-time period of that historical time period; each average ratio is calculated in the following way: Obtain the historical sub-sales data corresponding to each of the aforementioned historical sub-time periods; Based on each of the historical sub-sales data and each of the historical total sales data, the proportion of historical sub-sales data corresponding to each of the historical sub-time periods is obtained; Based on the proportion of each of the historical sub-sales data, determine the average proportion corresponding to each of the historical sub-time periods; The step of generating labor demand information corresponding to each predicted sub-time period based on each predicted sub-sales data includes: Obtain the target historical smart purchase sales ratio for each of the aforementioned historical sub-time periods; Based on each predicted sub-sales data and each target historical smart purchase sales ratio, determine the manual bank sales data corresponding to each predicted sub-time period. Based on the sales data of each of the aforementioned manual banking counters, labor demand information corresponding to each of the aforementioned predicted sub-time periods is generated.

2. The method according to claim 1, characterized in that, The step of obtaining the target historical smart purchase volume ratio corresponding to each of the historical sub-time periods includes: Obtain the historical intelligent purchase and sales data corresponding to each of the aforementioned historical sub-time periods; Based on each of the historical smart purchase sales data and each of the historical sub-sales data, the historical smart purchase sales ratio corresponding to each of the historical sub-time periods is obtained; Based on each historical smart purchase sales ratio, obtain the target historical smart purchase sales ratio corresponding to each of the historical sub-time periods.

3. The method according to claim 1, characterized in that, The step of determining the manual bank sales data corresponding to each predicted sub-period based on each predicted sub-sales data and each target historical smart purchase sales ratio includes: Based on each predicted sub-sales data and each target historical smart purchase sales ratio, the predicted smart purchase sales data corresponding to each predicted sub-time period is determined. Calculate the difference between each predicted sub-sales data and each predicted intelligent purchase sales data to obtain the manual bank sales data corresponding to each predicted sub-time period.

4. The method according to claim 1, characterized in that, The method further includes: Based on pre-set evaluation indicators, the rationality of the employment demand information corresponding to each predicted sub-time period is evaluated.

5. The method according to claim 4, characterized in that, in, The evaluation indicators include the average number of orders processed per target unit working hour and the target coefficient of variation; the step of evaluating the rationality of the labor demand information corresponding to each predicted sub-time period based on the pre-set evaluation indicators includes: Obtain the labor demand information corresponding to each of the predicted sub-time periods and the actual labor sales data corresponding to each of the predicted time periods; Based on the labor demand information corresponding to each of the predicted sub-time periods, obtain the total labor demand information corresponding to each of the predicted time periods. Based on the actual sales data of each manual bank counter and the total labor demand information for each, determine the order volume per unit working hour for each predicted time period; Based on the order volume per unit working hour corresponding to each predicted time period, determine the average order volume per unit working hour; Based on the average number of orders processed per unit of working hours and the number of orders processed per unit of working hours corresponding to each of the predicted time periods, the standard deviation of the number of orders processed per unit of working hours is determined. The coefficient of variation is obtained by calculating the ratio of the standard deviation of the number of orders processed per unit of working hours to the mean of the number of orders processed per unit of working hours. If the average number of orders processed per unit of working hours is greater than the average number of orders processed per target unit of working hours and the coefficient of variation is less than the target coefficient of variation, then the labor demand information is determined to be reasonable.

6. A labor demand forecasting device, characterized in that, The apparatus is applied to the labor demand forecasting method as described in any one of claims 1-5; the apparatus comprises: The building module is used to construct a time series decomposition model based on multiple historical total sales data corresponding to multiple historical time periods obtained in advance. The first generation module is used to generate multiple predicted total sales data corresponding to multiple predicted time periods based on the time series decomposition model. The segmentation module is used to divide each predicted total sales data into multiple predicted sub-sales data corresponding to each predicted sub-time period, based on the predicted total sales data and multiple pre-acquired average ratios. The second generation module is used to generate employment demand information corresponding to each of the predicted sub-time periods based on the predicted sub-sales data.

7. An electronic device, characterized in that, The electronic device includes a processor and a memory, the memory storing computer-executable instructions executable by the processor, the processor executing the computer-executable instructions to implement the method of any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions that, when invoked and executed by a processor, cause the processor to perform the method described in any one of claims 1 to 5.

Images