Replenishment Method, Device and Electronic Equipment

Abstract

The present disclosure relates to a replenishment method, apparatus and electronic device, and relates to the technical field of inventory management. The replenishment method includes: obtaining historical sales and inventory data sequences of multiple sales points with different out-of-stock types within a first preset time period; inputting the historical sales and inventory data sequences of each sales point into replenishment prediction models corresponding to different out-of-stock types to obtain evaluation values of multiple sales points, where the multiple out-of-stock types at least include an out-of-stock type and an impending out-of-stock type, and the replenishment prediction models corresponding to the out-of-stock type and the impending out-of-stock type are both binary classification models; determining a replenishment priority sequence of multiple sales points according to the evaluation values; and replenishing the multiple sales points according to the replenishment priority sequence. Instead of directly predicting the replenishment quantity of each sales point, the present disclosure determines the sales points that need to be replenished first according to the determined replenishment priority sequence, is not prone to overfitting, can simply and effectively estimate the sales points that are more urgently in need of replenishment, and replenish these sales points.

Description

Technical Field

[0001] The present disclosure relates to the technical field of inventory management, and in particular, to a replenishment method, apparatus, and electronic device. Background Art

[0002] The replenishment strategies for electronic products vary greatly during different life cycles. Due to the long production cycle of electronic products, during the first and middle stages of production capacity ramping up, the stockpiling is usually insufficient, and there will be situations where products are out of stock but cannot be replenished in time.

[0003] Currently, for the replenishment schemes of other types of products, based on the store inventory and sales situation of each sales point, combined with product information and sales point portrait features, statistical methods such as AR (Auto Regressive), MA (Moving Average), and ARIMA (Auto Regressive Integrated Moving Average) in statistics, or deep learning models such as RNN and CNN are used to fit a regression model to predict the demand for products, and corresponding replenishment is carried out in combination with the available sales days.

[0004] For electronic products, usually with a high unit price, the sales sequence of a single sales point is very sparse on a daily basis, and the number of days of retail sales is relatively large. Training a complex machine learning model is prone to overfitting and has low accuracy. Summary of the Invention

[0005] To overcome the problems existing in the related art, the present disclosure provides a replenishment method, apparatus, and electronic device.

[0006] According to a first aspect of an embodiment of the present disclosure, a replenishment method is provided, and the method includes:

[0007] Obtain historical sales and inventory data sequences of multiple sales points with different out-of-stock types within a first preset time period;

[0008] Input the historical sales and inventory data sequence of each sales point into a replenishment prediction model corresponding to different out-of-stock types to obtain evaluation values of the multiple sales points. The multiple out-of-stock types at least include an out-of-stock type and an impending out-of-stock type, and the replenishment prediction models corresponding to the out-of-stock type and the impending out-of-stock type are both binary classification models;

[0009] Determine a replenishment priority sequence of the multiple sales points according to the evaluation values;

[0010] Carry out replenishment for the multiple sales points according to the replenishment priority sequence.

[0011] Optionally, the carrying out replenishment for the multiple sales points according to the replenishment priority sequence includes:

[0012] Taking the target sales points to be replenished as including the top K sales points in the replenishment priority sequence and the total quantity of replenishment for the target sales points being less than or equal to the total quantity of commodity inventory as constraints, and aiming to maximize the target revenue value within a second preset duration after replenishing the target sales points, linear programming is performed to obtain the replenishment quantity for each of the target sales points, where the target revenue value is related to the sales rate of the target sales points within the second preset duration and the proportion of the sales points with inventory after the second preset duration among all sales points.

[0013] Optionally, the objective function of the linear programming is:

[0014] score = w 1 a + w 2 b

[0015] where score is the target revenue value, w 1 and w 2 are hyperparameters, a is the sales rate of the target sales points within the second preset duration, and b is the proportion of the sales points with inventory after the second preset duration among all sales points.

[0016] Optionally, taking the target sales points to be replenished as including the top K sales points in the replenishment priority sequence and the total quantity of replenishment for the target sales points being less than or equal to the total quantity of commodity inventory as constraints, and aiming to maximize the target revenue value within a second preset duration after replenishing the target sales points, linear programming is performed to obtain the replenishment quantity for each of the target sales points, including:

[0017] Obtain an initial coverage rate, where the coverage rate represents the proportion of the target sales points among the multiple sales points;

[0018] Determine the top K sales points from the replenishment priority sequence according to the coverage rate;

[0019] According to the total quantity of commodity inventory, determine each replenishment sequence for the top K sales points, where one replenishment sequence includes candidate replenishment quantities for each of the top K sales points;

[0020] For each replenishment sequence, determine the target revenue value within the second preset duration after replenishing the top K sales points according to the replenishment sequence;

[0021] In the case where there is a target revenue value greater than a preset threshold among the target revenue values corresponding to each replenishment sequence, determine the candidate replenishment quantities in the replenishment sequence corresponding to the maximum target revenue value as the target replenishment quantities for the top K sales points.

[0022] Optionally, the method further includes:

[0023] After replenishing the top K sales points according to the target replenishment quantity, if the target revenue value within the second preset duration is less than the reference revenue value, adjust the value of the coverage rate, and return to execute the step of determining the top K sales points from the replenishment priority sequence according to the adjusted coverage rate.

[0024] Optionally, the out-of-stock types further include the normal in-sale type. The current store inventory data of the to-be-supplemented goods at the out-of-stock type sales points is equal to zero. The current store inventory data of the to-be-supplemented goods at the soon-to-be-out-of-stock type sales points is greater than zero and less than or equal to a first threshold. The current store inventory data of the to-be-supplemented goods at the normal in-sale type sales points is greater than the first threshold and the historical sales volume is less than a second threshold. The evaluation values include a first evaluation value, a second evaluation value, and a third evaluation value;

[0025] The step of inputting the historical sales and inventory data sequences of each sales point into the replenishment prediction models corresponding to different out-of-stock types to obtain the evaluation values of the multiple sales points includes:

[0026] Input the historical sales and inventory data sequence of the to-be-supplemented goods at the out-of-stock type sales points into the first binary classification prediction model to obtain the first evaluation value;

[0027] Input the historical sales and inventory data sequence of the to-be-supplemented goods at the soon-to-be-out-of-stock type sales points into the second binary classification prediction model to obtain the second evaluation value;

[0028] Input the historical sales and inventory data sequence of the to-be-supplemented goods at the normal in-sale type sales points into the third prediction model to obtain the third evaluation value, where the third prediction model is an average value model.

[0029] Optionally, the training method of the first binary classification prediction model includes the following steps:

[0030] Obtain a first training sample set, where the first training sample set includes the first sales feature data of the sales points where the to-be-supplemented goods are characterized as out-of-stock type in the historical state within the third preset duration after replenishment. The first sales feature data includes first sales volume data. The first training sample set includes a first positive sample set and a first negative sample set. The first positive sample set includes the first sales feature data of the sales points where the first sales volume data is greater than a third threshold. The first negative sample set includes the first sales feature data of the sales points where the first sales volume data is less than or equal to the third threshold;

[0031] Input the first training sample set into the initial first binary classification prediction model, and use the cross-entropy loss function to iteratively train the initial first prediction model to obtain the first binary classification prediction model.

[0032] Optionally, the training method of the second binary classification prediction model includes the following steps:

[0033] Obtain a second training sample set, where the second training sample set includes the second sales feature data within a third preset duration after replenishment at a sales point where the store inventory data of the product to be replenished is greater than zero and less than or equal to the first threshold in the historical state. The second sales feature data includes second sales volume data; the second training sample set includes a second positive sample set and a second negative sample set. The second positive sample set includes the second sales feature data of sales points with a difference data greater than a third threshold, and the second negative sample set includes the second sales feature data of sales points with the difference data less than or equal to the third threshold. The difference data is the difference between the store inventory data of the sales point before replenishment and the second sales data within the third preset duration after replenishment.

[0034] Input the second training sample set into the initial second binary classification prediction model, and use the cross-entropy loss function to iteratively train the initial second prediction model to obtain the second binary classification prediction model.

[0035] According to a second aspect of the embodiments of the present disclosure, there is provided a replenishment device, the device includes:

[0036] An acquisition module, configured to acquire historical sales and inventory data sequences of multiple sales points with different out-of-stock types within a first preset duration;

[0037] An obtaining module, configured to input the historical sales and inventory data sequences of each sales point into the replenishment prediction models corresponding to different out-of-stock types to obtain evaluation values of the multiple sales points. The multiple out-of-stock types at least include an out-of-stock type and an impending out-of-stock type, and the replenishment prediction models corresponding to the out-of-stock type and the impending out-of-stock type are both binary classification models;

[0038] A first determination module, configured to determine a replenishment priority sequence of the multiple sales points according to the evaluation values;

[0039] A second determination module, configured to replenish the multiple sales points according to the replenishment priority sequence.

[0040] According to a third aspect of the embodiments of the present disclosure, there is provided an electronic device, the electronic device includes:

[0041] A first memory and a first processor;

[0042] The first memory is used to store program code and transmit the program code to the first processor;

[0043] The first processor is used to execute the method described in the first aspect of the embodiments of the present disclosure according to the instructions in the program code.

[0044] In the present disclosure, the historical sales and inventory data sequences of each sales point are input into replenishment prediction models corresponding to different out-of-stock types to obtain evaluation values of multiple sales points, and then a replenishment priority sequence of multiple sales points is determined according to the evaluation values to replenish the multiple sales points. In the related art, for products with relatively dense sales sequences, a deep learning model is generally used to fit a regression model to predict the demand for goods, while for products with relatively sparse sales sequences, overfitting is likely to occur. Therefore, the replenishment prediction models corresponding to the out-of-stock types and the upcoming out-of-stock types in the present disclosure adopt a binary classification model, which does not directly predict the replenishment quantity of each sales point, but determines the sales points to be replenished first according to the determined replenishment priority sequence, and is not prone to overfitting. It can simply and effectively estimate the sales points that are more urgently in need of replenishment and replenish these sales points. At the same time, in the related art, a fixed replenishment value is predicted for each sales point, and there is a situation where multiple sales points require the same quantity but have significantly different degrees of urgency. The replenishment priority sequence predicted based on the binary classification model in the present disclosure can determine the urgency of replenishment for each sales point, making the present disclosure have good interpretability.

[0045] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, showing embodiments consistent with the present disclosure and used together with the specification to explain the principles of the present disclosure.

[0047] Figure 1 is a flowchart of a replenishment method shown according to an exemplary embodiment.

[0048] Figure 2 is a block diagram of a replenishment device shown according to an exemplary embodiment.

[0049] Figure 3 is a block diagram of a device for replenishment shown according to an exemplary embodiment.

[0050] Figure 4 is a block diagram of a device for implementing a replenishment method shown according to an exemplary embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0051] Exemplary embodiments will be described in detail herein, and examples thereof are shown in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0052] In the related art, electronic products usually exhibit the characteristics of high unit price, rapid technological upgrading, and short life cycle. Currently, the replenishment at offline sales points relies heavily on manual labor. Once the supply or demand for products is inaccurately grasped, it is easy to cause overstocking, increased loss rate, and rising costs; if the replenishment is insufficient, the in-stock rate will decrease, the products will be out of stock, thus affecting sales. However, the method of estimating sales volume and replenishing goods for each sales point does not take into account the urgency of out-of-stock at the sales point. The sales capabilities of different sales points vary greatly. If all the goods are given to the sales points with strong sales capabilities, some sales points with weak sales capabilities will be out of stock for a long time, directly affecting the store's sales; if all out-of-stock sales points try to replenish goods, first, the supply chain warehouse inventory is limited and full replenishment cannot be achieved, and second, replenishing more goods at sales points with weak sales capabilities will reduce the overall turnover speed of the products. In the case of limited supply, how much warehouse inventory is used for emergency replenishment of out-of-stock sales points and how much is used for normal replenishment of sales points with faster sales become common problems in intelligent replenishment under tight supply chain conditions.

[0053] Figure 1 is a flowchart of a replenishment method shown according to an exemplary embodiment, which can be used for products with relatively sparse sales sequences, such as Figure 1 shown, the replenishment method includes the following steps.

[0054] In step S101, obtain the historical sales and inventory data sequences of multiple sales points with different out-of-stock types within a first preset time period.

[0055] Exemplarily, the first preset time period is a relatively long time period set in advance, which can be 5 days, 7 days, 10 days, etc. Different out-of-stock types can include the out-of-stock type and the about-to-be-out-of-stock type. Among them, for the sales points of the out-of-stock type, the current store inventory data of the goods to be replenished is equal to zero, and for the sales points of the about-to-be-out-of-stock type, the current store inventory data of the goods to be replenished is greater than zero and less than or equal to a first threshold. Different out-of-stock types can also include the normal on-sale type, and for the sales points of the normal on-sale type, the current store inventory data of the goods to be replenished is greater than the first threshold and the historical sales volume is less than a second threshold.

[0056] Exemplarily, different out-of-stock types can also be defined by different sales gaps for each sales point. For example, they are divided into a first type and a second type. The demand for goods at the sales points corresponding to the first type is greater than 0 and less than or equal to a third threshold, and the demand for goods at the sales points corresponding to the second type is greater than the third threshold.

[0057] Exemplarily, the historical sales and inventory data sequence can include the data sequence of the sales points corresponding to different out-of-stock types within a first preset duration in history and the data sequence of the inventory of goods at the sales points after sales and / or replenishment. For example, within the first five days before the current time, the historical sales and inventory data sequence includes a sales data sequence and an inventory data sequence. The sales data sequence is 13012, and the inventory data sequence is 40321. It can be understood that the replenishment quantity on the third day is 3, and there is no replenishment at other times.

[0058] In step S102, the historical sales and inventory data sequence of each sales point is input into the replenishment prediction models corresponding to different out-of-stock types, and evaluation values of multiple sales points are obtained. The multiple out-of-stock types at least include the out-of-stock type and the impending out-of-stock type. The replenishment prediction models corresponding to the out-of-stock type and the impending out-of-stock type are both binary classification models.

[0059] Exemplarily, the replenishment prediction model is used to predict the urgency of out-of-stock at the sales points corresponding to each out-of-stock type according to the input sales and inventory data sequence. The evaluation value can represent the out-of-stock tightness of each sales point. The higher the evaluation value, the higher the urgency of out-of-stock. The binary classification model is a two-value classification model, and the prediction result can be formally expressed as 0 or 1. Generally, the numbers output by the binary classification model are not directly the numbers represented by 0 or 1, but the probabilities of the input data being a certain situation. For example, if the evaluation value of the binary classification model is 0.8, the corresponding output value after rounding is 1. Or in another case, the binary classification model outputs the evaluation value greater than the valuation threshold as 1. Then, when the evaluation value of the binary classification model is 0.8 and the valuation threshold is 0.85, the corresponding output value is 0. In the present disclosure, the binary classification model does not output 0 or 1, but outputs the corresponding evaluation value, so as to sort the out-of-stock urgency of the sales points according to the evaluation values of each sales point.

[0060] Exemplarily, each out-of-stock type in different out-of-stock types corresponds to a replenishment prediction model. Among them, the replenishment prediction models corresponding to the out-of-stock type and the impending out-of-stock type are both binary classification models. After inputting the historical sales and inventory data sequences corresponding to the out-of-stock type and the impending out-of-stock type into the corresponding replenishment prediction models, the evaluation values corresponding to each sales point can be output.

[0061] In step S103, according to the evaluation values, a replenishment priority sequence of multiple sales points is determined.

[0062] Exemplarily, the replenishment priority sequence is the result of sorting the out-of-stock urgency levels of different sales points. The evaluation values output for the out-of-stock types and the about-to-be-out-of-stock types are sorted by priority to obtain the replenishment priority sequences of multiple sales points.

[0063] In one implementation, the evaluation values output for the out-of-stock types and the about-to-be-out-of-stock types are sorted to obtain the replenishment priority sequence. Additionally, since the training conditions of the replenishment prediction models corresponding to each out-of-stock type or the parameters of the objective functions of the models are different, in another implementation, based on the replenishment prediction model corresponding to the out-of-stock type and the output first evaluation value sequence, and based on the second evaluation value sequence output by the replenishment prediction model corresponding to the about-to-be-out-of-stock type and a preset ratio value, the replenishment priority sequences of multiple sales points are determined.

[0064] Exemplarily, if the first evaluation value sequence corresponding to the out-of-stock type is 0.95, 0.9, 0.88, 0.8, the second evaluation value sequence corresponding to the about-to-be-out-of-stock type is 0.92, 0.89, 0.85, 0.82, and the preset ratio value is 0.9. According to the sorting method of the above one implementation, the replenishment priority sequence is 0.95, 0.92, 0.9, 0.89, 0.88, 0.85, 0.82, 0.8. According to the sorting method of the above another implementation, the replenishment priority sequence is 0.95, 0.9, 0.88, 0.92, 0.89, 0.8, 0.85, 0.82.

[0065] In step 104, multiple sales points are replenished according to the replenishment priority sequence.

[0066] Exemplarily, multiple sales points are replenished according to the replenishment priority sequence. Among them, a K value needs to be determined first to replenish the first K sales points in the replenishment priority sequence. Generally, the replenishment priority sequence output by the binary classification model cannot predict the actual replenishment quantity required for each sales point, but the replenishment quantity for the first K sales points is determined based on the replenishment priority sequence and the demand quantities corresponding to the first K sales points. Among them, the out-of-stock quantity can be determined according to the average value of the sales volume of each sales point within the first preset time period, or can be determined according to the replenishment quantity applied by the sales point.

[0067] The present disclosure inputs the historical sales and inventory data sequences of each point of sale into replenishment prediction models corresponding to different out-of-stock types to obtain evaluation values of multiple points of sale, and thus determines a replenishment priority sequence of multiple points of sale according to the evaluation values, and replenishes the multiple points of sale. In the related art, for products with relatively dense sales sequences, a deep learning model is generally used to fit a regression model to predict the demand for goods, while for products with relatively sparse sales sequences, overfitting is likely to occur. Therefore, the replenishment prediction models corresponding to the out-of-stock types and the upcoming out-of-stock types in the present disclosure adopt a binary classification model, which does not directly predict the replenishment quantity of each point of sale, but determines the points of sale to be replenished first according to the determined replenishment priority sequence, and is not prone to overfitting. It can simply and effectively estimate the points of sale that are more urgently in need of replenishment and replenish these points of sale. At the same time, in the related art, a fixed replenishment value is predicted for each point of sale, and there is a situation where multiple points of sale require the same quantity but have significantly different urgencies. The replenishment priority sequence predicted based on the binary classification model in the present disclosure can determine the replenishment urgency of each point of sale, making the present disclosure have good interpretability.

[0068] In some embodiments, replenishing multiple points of sale according to the replenishment priority sequence includes:

[0069] Taking the target points of sale to be replenished including the first K points of sale in the replenishment priority sequence and the total quantity of replenishing the target points of sale being less than or equal to the total quantity of commodity inventory as a constraint, and taking maximizing the target revenue value within a second preset duration after replenishing the target points of sale as a goal, performing linear programming to obtain the replenishment quantity of each target point of sale, where the target revenue value is related to the sales rate of the target points of sale within the second preset duration and the proportion of the points of sale with inventory after the second preset duration among all points of sale.

[0070] Exemplarily, the value of K is less than the total number of points of sale, where K is generally a preset value. The second preset duration can be a duration equal to the first preset duration. The target revenue value is determined within the second preset duration after replenishing the first K points of sale. Linear programming (abbreviated as LP), is an important branch in operations research that has been studied earlier, developed faster, widely applied, and has relatively mature methods. It can assist in scientific management and optimal decision-making, and is a mathematical method for studying the extreme value problem of a linear objective function under linear constraint conditions. The linear programming method can adopt a multi-objective integer programming method, that is, both input and output variables of the multi-objective integer programming take integers, and the point where the objective function takes the maximum or minimum value in the feasible region. Adopting the multi-objective integer programming method for the first K points of sale in the replenishment priority sequence can determine the optimal replenishment plan.

[0071] Exemplarily, the sales rate of a target sales point within a second preset duration refers to the ratio of the quantity sold to the quantity replenished within the second preset duration after replenishing multiple target sales points. The proportion of sales points with inventory after the second preset duration among all sales points refers to the proportion of the number of sales points with inventory after sales within the second preset duration after replenishing multiple target sales points among all sales points.

[0072] In some embodiments, the objective function of linear programming is:

[0073] score = w 1 a + w 2 b

[0074] where score is the target revenue value, w 1 and w 2 are hyperparameters, a is the sales rate of the target sales point within the second preset duration, and b is the proportion of sales points with inventory after the second preset duration among all sales points.

[0075] Exemplarily, the hyperparameters can be determined according to the sales requirements in different stages of the business. For example, in the case of high sales volume business requirements, the ratio of w 1 to w 2 can be set to be relatively small, so that the sales points with more sales volume are replenished preferentially. The value of a will increase, the value of b will decrease, and the value of score will also increase. In the case of business requirements to meet replenishment, more replenishment needs to be given to out-of-stock sales points. The value of a will decrease, and the value of b will increase. By optimizing this indicator, a balance can be achieved between the turnover speed of replenishment and meeting the needs of most replenished sales points.

[0076] In some embodiments, with the constraint that the target sales points for replenishment include the first K sales points in the replenishment priority sequence and the total quantity of replenishment for the target sales points is less than or equal to the total quantity of commodity inventory, and with the goal of maximizing the target revenue value within the second preset duration after replenishing the target sales points, linear programming is performed to obtain the replenishment quantity for each target sales point, including:

[0077] Obtain the initial coverage rate, where the coverage rate represents the proportion of target sales points among multiple sales points;

[0078] Determine the first K sales points from the replenishment priority sequence according to the coverage rate;

[0079] According to the total quantity of commodity inventory, determine each replenishment sequence for the first K sales points. One replenishment sequence includes candidate replenishment quantities for each of the first K sales points;

[0080] For each replenishment sequence, determine the target revenue value within the second preset time period after replenishing the first K sales points according to the replenishment sequence.

[0081] When there is a target revenue value greater than the preset threshold among the target revenue values corresponding to each replenishment sequence, determine each candidate replenishment quantity in the replenishment sequence corresponding to the maximum target revenue value as the target replenishment quantity for the first K sales points.

[0082] Exemplarily, an initial coverage rate can determine an initial value according to a preset upper limit and a preset lower limit of the coverage rate, which can be the average of the preset upper limit and the preset lower limit of the coverage rate. The value of K for the first K sales points can be determined according to the product of the coverage rate and the total number of sales points to be replenished. A replenishment sequence refers to various different replenishment methods for the first K sales points in the replenishment priority sequence.

[0083] For example, when the total quantity of commodity inventory is 4 and the value of K is 3, the replenishment sequence can include 211 or 111. In the case where the replenishment sequence is 211, it can be understood that the higher the urgency of replenishment at the position closer to the front of the replenishment priority sequence, so the more the replenishment quantity at the position closer to the front of the replenishment priority sequence. In the case where the replenishment sequence is 111, it can be understood that the replenishment requirements of the first 3 sales points can be preferentially satisfied. If there is remaining inventory, the remaining inventory can be used to replenish the 4th sales point and subsequent sales points, and the replenishment method for the sales points after the first K sales points is not limited in this disclosure. In another embodiment, in the case where the replenishment sequence is 111, all the above numbers do not represent actual quantities, but the satisfaction degrees of commodity replenishment. That is, when the replenishment requirement values of the first 3 sales points are 10, 15, and 12, and the commodity inventory is sufficient, the actual replenishment values for the first 3 sales points are 10, 15, and 12. If there is remaining inventory, the remaining inventory can be used to replenish the 4th sales point and subsequent sales points.

[0084] For each replenishment sequence, the target revenue value of the objective function can be calculated according to the above objective function. The preset threshold is a value preset according to the demand and can be determined according to the target sales volume. When there is a target revenue value greater than the preset threshold, each candidate replenishment quantity in the replenishment sequence corresponding to the maximum target revenue value can be determined as the target replenishment quantity for the first K sales points.

[0085] In some embodiments, the method further includes:

[0086] After replenishing the top K sales points according to the target replenishment quantity, if the target revenue value within the second preset duration is less than the reference revenue value, then adjust the value of the coverage rate, and return to execute the step of determining the top K sales points from the replenishment priority sequence according to the adjusted coverage rate.

[0087] Exemplarily, the target revenue value refers to the actual revenue situation after replenishing the top K sales points according to the target replenishment quantity. The reference revenue value is a value determined according to the above-mentioned preset threshold, and can be determined according to the preset threshold, the maximum inventory backlog rate, and the inventory transfer ability of the sales points. According to the target revenue value corresponding to the actual revenue situation, determine whether the target revenue value is less than the reference revenue value. If it is less than the reference revenue value, then adjust the value of the coverage rate.

[0088] In some embodiments, the out-of-stock types further include the normal on-sale type. The current store inventory data of the to-be-supplemented goods at the out-of-stock type sales points is equal to zero. The current store inventory data of the to-be-supplemented goods at the about-to-be-out-of-stock type sales points is greater than zero and less than or equal to the first threshold. The current store inventory data of the to-be-supplemented goods at the normal on-sale type sales points is greater than the first threshold and the historical sales volume is less than the second threshold. The evaluation values include the first evaluation value, the second evaluation value, and the third evaluation value;

[0089] The step of inputting the historical sales and inventory data sequences of each sales point into the replenishment prediction models corresponding to different out-of-stock types to obtain the evaluation values of multiple sales points includes:

[0090] Input the historical sales and inventory data sequence of the to-be-supplemented goods at the out-of-stock type sales points into the first binary classification prediction model to obtain the first evaluation value;

[0091] Input the historical sales and inventory data sequence of the to-be-supplemented goods at the about-to-be-out-of-stock type sales points into the second binary classification prediction model to obtain the second evaluation value;

[0092] Input the historical sales and inventory data sequence of the to-be-supplemented goods at the normal on-sale type sales points into the third prediction model to obtain the third evaluation value, where the third prediction model is an average value model.

[0093] Exemplarily, the replenishment prediction models corresponding to the out-of-stock type sales points and the about-to-be-out-of-stock type sales points are both binary classification models, and the replenishment prediction model corresponding to the normal on-sale type sales points is an average value model. According to the replenishment prediction models corresponding to different out-of-stock types, the evaluation values corresponding to multiple sales points can be obtained.

[0094] Among them, the binary classification model may include, but is not limited to, binary classification models such as LightGBM model, DNN model, CNN model, RNN model, etc. For the LightGBM model, the historical sales and inventory data sequence of each sales point needs to pass through sliding windows of different sizes to extract multiple segmented data. For the CNN model, time series features can be extracted through a one-dimensional convolutional layer. The RNN model itself is suitable for time series feature input.

[0095] Exemplarily, the loss function of the binary classification model can adopt the cross-entropy loss function. After adjusting the parameters of the binary classification model to minimize the loss value of the cross-entropy loss function, the data of the second preset duration after the training set can be used to evaluate the positive sample accuracy of the K value, measure the model effect under different business requirements, and finally select the corresponding category of the binary classification model. Among them, the positive sample is the target sales point with sales volume after replenishing the target sales point, and the negative sample is the target sales point with a sales volume of 0 after replenishing the target sales point. In addition, since replenishment may be carried out for sales points that are already out of stock but have relatively small historical sales data, in actual situations, after replenishment, this store will not generate sales volume, and such stores are negative samples.

[0096] In some embodiments, the training method of the first binary classification prediction model includes the following steps:

[0097] Obtain a first training sample set. The first training sample set includes the first sales feature data of sales points where the goods to be replenished were characterized as out-of-stock types in the historical state within the third preset duration after replenishment. The first sales feature data includes the first sales volume data. The first training sample set includes a first positive sample set and a first negative sample set. The first positive sample set includes the first sales feature data of sales points where the first sales volume data is greater than the third threshold, and the first negative sample set includes the first sales feature data of sales points where the first sales volume data is less than or equal to the third threshold;

[0098] Input the first training sample set into the initial first binary classification prediction model, and use the cross-entropy loss function to iteratively train the initial first prediction model to obtain the first binary classification prediction model.

[0099] Exemplarily, for the first training sample set of the first binary classification prediction model, data such as the first sales data, store inventory data, in-transit data, store portrait, product portrait, promotion features, holiday features, etc. can be input into the initial first binary classification prediction model after feature engineering to iteratively train the initial first binary classification prediction model to obtain the trained first binary classification prediction model. The first sales data refers to the data sold by the point of sale of the out-of-stock type in the third preset duration after replenishment in the historical state. The store inventory data is the remaining quantity of the out-of-stock type of point of sale in the third preset duration after replenishment in the historical state. The in-transit data is the quantity during the transportation of the goods when replenishing the point of sale. The store portrait can characterize features such as the pedestrian flow in the city where the point of sale is located, the advantages and disadvantages of the geographical location, the decoration level, and the number of store clerks. The product portrait can characterize the performance of the product to be replenished and / or the population demand characteristics of the product appearance. The promotion features can characterize features such as promotion activities during new product launches or before the next new product launch or holiday promotion activities. The holiday features can characterize the feature of increased pedestrian flow during holidays.

[0100] Exemplarily, the third threshold can be 0, that is, the data of the point of sale with the first sales data greater than 0 is used as the first positive sample set, and the data of the point of sale with the first sales data less than 0 is used as the first negative sample set.

[0101] In some embodiments, the training method of the second binary classification prediction model includes the following steps:

[0102] Obtain a second training sample set, which includes the second sales feature data of the point of sale that replenishes the goods when the store inventory data of the goods to be replenished in the historical state is greater than zero and less than or equal to the first threshold within the third preset duration after replenishment. The second sales feature data includes the second sales data; the second training sample set includes a second positive sample set and a second negative sample set. The second positive sample set includes the second sales feature data of the point of sale with the difference data greater than the third threshold, and the second negative sample set includes the second sales feature data of the point of sale with the difference data less than or equal to the third threshold. The difference data is the difference between the store inventory data of the point of sale before replenishment and the second sales data within the third preset duration after replenishment;

[0103] Input the second training sample set into the initial second binary classification prediction model, and use the cross-entropy loss function to iteratively train the initial second prediction model to obtain the second binary classification prediction model.

[0104] Exemplarily, for the second training sample set of the second binary classification prediction model, the data used is similar to that of the first binary classification prediction model, which will not be elaborated here. The differential data is the difference between the store inventory data when there is no replenishment and the second sales data within the third preset duration after replenishment. Generally, the third threshold can be set to 0, that is, the data of the sales points with differential data greater than 0 is used as the second positive sample set, and the data of the sales points with differential data less than 0 is used as the second negative sample set.

[0105] In some embodiments, the third prediction model is an average value model.

[0106] Exemplarily, for the normal on-sale type, if the historical sales data sequence is relatively sparse, the average value model is adopted; if the historical sales data sequence is relatively dense, a regression model fitted by a deep learning model in the related art is used to predict the replenishment quantity of the demand for goods, which is not limited here.

[0107] Exemplarily, the historical sales and inventory data sequence is input into the average value model. After the average value preprocesses the historical sales and inventory data, the average value of the preprocessed data is used as the third evaluation value. It can be understood that in any case, among the replenishment priority sequences, the sales points of the normal on-sale type are all behind the out-of-stock type and the soon-to-be-out-of-stock type.

[0108] Exemplarily, during the preprocessing process, the data that is significantly different from the rest of the data is removed, and the removed data is regarded as a missing value, and the data of the day before the missing value is used to fill the missing value. If the data of the day before the missing value is still significantly different from the rest of the data, there is no need to fill the missing value, and the average value of the rest of the data is used as the third evaluation value. For example, in the sales sequence from the 10th to the 14th of a certain month, it is 93243. The situation where the sales volume on the 10th is particularly large may be due to the promotion stage or holiday stage at that time, so the sales volume data of 9 on the 10th is removed. If the data on the 9th is still significantly different from the data on the 11th - 14th, the average value of 3 of the data on the 11th - 14th is used as the third evaluation value.

[0109] It can be understood that the third evaluation value can be used to determine the replenishment quantity required for the sales points of the normal on-sale type. That is, in the above example, the third evaluation value of 3 is used as the demand replenishment quantity for this sales point.

[0110] Exemplarily, for the first K sales points, if the value of K is 6, and the numbers of the determined out-of-stock type sales points, soon-to-be-out-of-stock type sales points, and normal on-sale type sales points are 3, 2, and 1 respectively. In this case, replenishment is normally carried out for the above 6 sales points. If the numbers of the out-of-stock type sales points and soon-to-be-out-of-stock type sales points are 4 and 2 respectively, the demand replenishment quantity of the normal on-sale type does not need to be considered.

[0111] Figure 2 is a block diagram of a replenishment device 120 shown according to an exemplary embodiment. Referring to Figure 2 , the device includes an acquisition module 121, an obtaining module 122, a first determination module 123, and a second determination module 124.

[0112] The acquisition module 121 is configured to acquire a historical sales and inventory data sequence of multiple sales points of different out-of-stock types within a first preset time period;

[0113] The obtaining module 122 is configured to input the historical sales and inventory data sequence of each sales point into a replenishment prediction model corresponding to a different out-of-stock type to obtain evaluation values of multiple sales points. The multiple out-of-stock types at least include an out-of-stock type and an impending out-of-stock type. The replenishment prediction models corresponding to the out-of-stock type and the impending out-of-stock type are both binary classification models;

[0114] The first determination module 123 is configured to determine a replenishment priority sequence of multiple sales points according to the evaluation values;

[0115] The second determination module 124 is configured to replenish multiple sales points according to the replenishment priority sequence.

[0116] In some embodiments, the second determination module 124 includes:

[0117] An obtaining sub-module, configured to perform linear programming with the constraint that the target sales points to be replenished include the first K sales points in the replenishment priority sequence and the total quantity of replenishing the target sales points is less than or equal to the total quantity of commodity inventory, and with the goal of maximizing the target revenue value within a second preset time period after replenishing the target sales points. The replenishment quantity of each target sales point is obtained. The target revenue value is related to the sales rate of the target sales point within the second preset time period and the proportion of the sales points with inventory after the second preset time period among all sales points.

[0118] In some embodiments, in the obtaining sub-module, the objective function of the linear programming is:

[0119] score = w 1 a + w 2 b

[0120] where score is the target revenue value, w 1 and w 2 are hyperparameters, a is the sales rate of the target sales point within the second preset time period, and b is the proportion of the sales points with inventory after the second preset time period among all sales points.

[0121] In some embodiments, the obtaining sub-module is configured to:

[0122] Obtain the initial coverage rate, where the coverage rate represents the proportion of the target vending point among multiple vending points;

[0123] Determine the top K vending points from the replenishment priority sequence according to the coverage rate;

[0124] According to the total quantity of commodity inventory, determine each replenishment sequence for the top K vending points, where a replenishment sequence includes candidate replenishment quantities for each vending point among the top K vending points;

[0125] For each replenishment sequence, determine the target revenue value within the second preset duration after replenishing the top K vending points according to the replenishment sequence;

[0126] When there is a target revenue value greater than the preset threshold among the target revenue values corresponding to each replenishment sequence, determine each candidate replenishment quantity in the replenishment sequence corresponding to the maximum target revenue value as the target replenishment quantity for the top K vending points.

[0127] In some embodiments, the replenishment device 120 further includes:

[0128] An adjustment module, configured to, after replenishing the top K vending points according to the target replenishment quantity, if the target revenue value within the second preset duration is less than the reference revenue value, adjust the value of the coverage rate, and return to execute the step of determining the top K vending points from the replenishment priority sequence according to the adjusted coverage rate.

[0129] In some embodiments, the out-of-stock types further include the normal on-sale type. The current store inventory data of the to-be-supplemented commodities at the out-of-stock vending points is equal to zero. The current store inventory data of the to-be-supplemented commodities at the about-to-be-out-of-stock vending points is greater than zero and less than or equal to the first threshold. The current store inventory data of the to-be-supplemented commodities at the normal on-sale vending points is greater than the first threshold and the historical sales volume is less than the second threshold. The evaluation values include the first evaluation value, the second evaluation value, and the third evaluation value;

[0130] The obtaining module 122 is configured to:

[0131] Input the historical sales and inventory data sequence of the to-be-supplemented commodities at the out-of-stock vending points into the first binary classification prediction model to obtain the first evaluation value;

[0132] Input the historical sales and inventory data sequence of the to-be-supplemented commodities at the about-to-be-out-of-stock vending points into the second binary classification prediction model to obtain the second evaluation value;

[0133] Input the historical sales and inventory data sequence of the to-be-supplemented commodities at the normal on-sale vending points into the third prediction model to obtain the third evaluation value, where the third prediction model is an average value model.

[0134] In some embodiments, in the obtaining module 122, the training method of the first binary classification prediction model includes the following steps:

[0135] Obtain a first training sample set, where the first training sample set includes first sales feature data of a sales point where the product to be replenished was characterized as out of stock in a historical state within a third preset duration after replenishment. The first sales feature data includes first sales volume data. The first training sample set includes a first positive sample set and a first negative sample set. The first positive sample set includes the first sales feature data of sales points where the first sales volume data is greater than a third threshold, and the first negative sample set includes the first sales feature data of sales points where the first sales volume data is less than or equal to the third threshold;

[0136] Input the first training sample set into an initial first binary classification prediction model, and use the cross-entropy loss function to iteratively train the initial first prediction model to obtain the first binary classification prediction model.

[0137] In some embodiments, in the obtaining module 122, the training method of the second binary classification prediction model includes the following steps:

[0138] Obtain a second training sample set, where the second training sample set includes second sales feature data of a sales point that replenishes the product when the in-store inventory data of the product to be replenished is greater than zero and less than or equal to a first threshold within a third preset duration after replenishment. The second sales feature data includes second sales volume data; the second training sample set includes a second positive sample set and a second negative sample set. The second positive sample set includes the second sales feature data of sales points where the difference data is greater than a third threshold, and the second negative sample set includes the second sales feature data of sales points where the difference data is less than or equal to the third threshold. The difference data is the difference between the in-store inventory data of the sales point before replenishment and the second sales data within the third preset duration after replenishment;

[0139] Input the second training sample set into an initial second binary classification prediction model, and use the cross-entropy loss function to iteratively train the initial second prediction model to obtain the second binary classification prediction model.

[0140] Regarding the device in the above embodiments, the specific manners in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0141] The present disclosure also provides a computer-readable storage medium, on which computer program instructions are stored. When the program instructions are executed by a third processor, the steps of the replenishment method provided by the present disclosure are implemented.

[0142] The present disclosure also provides an electronic device, which includes:

[0143] A first memory and a first processor;

[0144] The first memory is used to store program code and transfer the program code to the first processor;

[0145] The first processor is used to execute the method according to the first aspect of the embodiments of the present disclosure based on the instructions in the program code.

[0146] Figure 3 FIG. 8 is a block diagram of an apparatus 800 for replenishment according to an exemplary embodiment. For example, the apparatus 800 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0147] Referring to Figure 3 , the apparatus 800 may include one or more of the following components: a first processing component 802, a second memory 804, a first power supply component 806, a multimedia component 808, an audio component 810, a first input / output interface 812, a sensor component 814, and a communication component 816.

[0148] The first processing component 802 generally controls the overall operation of the apparatus 800, such as operations associated with display, telephone calls, data communication, camera operations, and recording operations. The first processing component 802 may include one or more second processors 820 to execute instructions to complete all or part of the steps of the above replenishment method. In addition, the first processing component 802 may include one or more modules to facilitate the interaction between the first processing component 802 and other components. For example, the first processing component 802 may include a multimedia module to facilitate the interaction between the multimedia component 808 and the first processing component 802.

[0149] The second memory 804 is configured to store various types of data to support the operation of the apparatus 800. Examples of these data include instructions for any application or method operating on the apparatus 800, contact data, phone book data, messages, pictures, videos, etc. The second memory 804 may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, a magnetic disk, or an optical disk.

[0150] The first power supply component 806 provides power to various components of the apparatus 800. The first power supply component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for the apparatus 800.

[0151] The multimedia component 808 includes a screen that provides an output interface between the device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen can be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors can sense not only the boundaries of touch or swipe actions but also detect the duration and pressure associated with the touch or swipe operation. In some embodiments, the multimedia component 808 includes a front camera and / or a rear camera. When the device 800 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera can receive external multimedia data. Each of the front camera and the rear camera can be a fixed optical lens system or have focal length and optical zoom capabilities.

[0152] The audio component 810 is configured to output and / or input audio signals. For example, the audio component 810 includes a microphone (MIC) that is configured to receive external audio signals when the device 800 is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signals can be further stored in the second memory 804 or transmitted via the communication component 816. In some embodiments, the audio component 810 further includes a speaker for outputting audio signals.

[0153] The first input / output interface 812 provides an interface between the first processing component 802 and a peripheral interface module, and the peripheral interface module can be a keyboard, a click wheel, buttons, etc. These buttons can include but are not limited to: a home button, a volume button, a power button, and a lock button.

[0154] The sensor component 814 includes one or more sensors for providing a status assessment of various aspects of the device 800. For example, the sensor component 814 can detect the on / off state of the device 800, the relative positioning of components, such as the display and the keypad of the device 800. The sensor component 814 can also detect a change in the position of the device 800 or a component of the device 800, the presence or absence of user contact with the device 800, the orientation or acceleration / deceleration of the device 800, and the temperature change of the device 800. The sensor component 814 can include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor component 814 can also include a light sensor, such as a CMOS or a CCD image sensor, for use in imaging applications. In some embodiments, the sensor component 814 can further include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0155] The communication component 816 is configured to facilitate communication, either wired or wirelessly, between the device 800 and other devices. The device 800 may access a wireless network based on communication standards, such as WiFi, 2G, or 3G, or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 further includes a Near Field Communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID) technology, Infrared Data Association (IrDA) technology, Ultra Wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0156] In an exemplary embodiment, the device 800 may be implemented by one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components for performing the above replenishment method.

[0157] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a second memory 804 including instructions, and the above instructions can be executed by a second processor 820 of the device 800 to complete the above replenishment method. For example, the non-transitory computer-readable storage medium may be a ROM, Random Access Memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0158] In another exemplary embodiment, a computer program product is also provided, and the computer program product includes a computer program capable of being executed by a programmable device, and the computer program has a code portion for performing the above replenishment method when executed by the programmable device.

[0159] Figure 4 is a block diagram of a device 1900 for implementing a replenishment method shown according to an exemplary embodiment. For example, the device 1900 may be provided as a server. Referring to Figure 4 , the device 1900 includes a second processing component 1922, which further includes one or more processors, and memory resources represented by a third memory 1932 for storing instructions executable by the second processing component 1922, such as application programs. The application programs stored in the third memory 1932 may include one or more modules each corresponding to a set of instructions. In addition, the second processing component 1922 is configured to execute instructions to perform the above replenishment method.

[0160] The apparatus 1900 may further include a second power component 1926 configured to perform power management of the apparatus 1900, a wired or wireless network interface 1950 configured to connect the apparatus 1900 to a network, and a second input / output interface 1958. The apparatus 1900 may operate based on an operating system stored in the third memory 1932, such as Windows Server TM , Mac OS X TM , Unix TM , Linux TM , FreeBSD TM or the like.

[0161] Other embodiments of the present disclosure will be readily apparent to those skilled in the art in view of the specification and practice of the present disclosure. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include known common knowledge or conventional technical means in the technical field not disclosed by the present disclosure. The specification and examples are only to be considered as exemplary, and the true scope and spirit of the present disclosure are pointed out by the claims.

[0162] It should be understood that the present disclosure is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present disclosure is only limited by the appended claims.

Claims

1. A replenishment method, characterized in that, the method includes: Obtaining historical sales and inventory data sequences of multiple sales points with different out-of-stock types within a first preset time period; Inputting the historical sales and inventory data sequences of each sales point into replenishment prediction models corresponding to different out-of-stock types to obtain evaluation values of the multiple sales points. The multiple out-of-stock types at least include the out-of-stock type and the soon-to-be-out-of-stock type. The replenishment prediction models corresponding to the out-of-stock type and the soon-to-be-out-of-stock type are both binary classification models; Determining a replenishment priority sequence of the multiple sales points according to the evaluation values; Replenishing the multiple sales points according to the replenishment priority sequence.

2. The method according to claim 1, characterized in that, the replenishing the multiple sales points according to the replenishment priority sequence includes: Taking the target sales points to be replenished including the first K sales points in the replenishment priority sequence and the total quantity of replenishing the target sales points being less than or equal to the total quantity of commodity inventory as constraints, and taking maximizing the target revenue value within a second preset time period after replenishing the target sales points as the goal, performing linear programming to obtain the replenishment quantity of each target sales point, where the target revenue value is related to the sales rate of the target sales point within the second preset time period and the proportion of the sales points with inventory after the second preset time period among all sales points.

3. The method according to claim 2, characterized in that, the objective function of the linear programming is: score = w 1 a + w 2 b where score is the target revenue value, w 1 and w 2 are hyperparameters, a is the sales rate of the target vending point within the second preset time period, and b is the proportion of vending points with inventory among all vending points after the second preset time period.

4. The method according to claim 2, characterized in that, the taking the target sales points to be replenished including the first K sales points in the replenishment priority sequence and the total quantity of replenishing the target sales points being less than or equal to the total quantity of commodity inventory as constraints, and taking maximizing the target revenue value within a second preset time period after replenishing the target sales points as the goal, performing linear programming to obtain the replenishment quantity of each target sales point includes: Obtaining an initial coverage rate, where the coverage rate represents the proportion of the target sales points among the multiple sales points; Determining the first K sales points from the replenishment priority sequence according to the coverage rate; Determining, according to the total quantity of commodity inventory, each replenishment sequence for the first K sales points, and a replenishment sequence includes candidate replenishment quantities for each of the first K sales points; For each replenishment sequence, determining the target revenue value within the second preset time period after replenishing the first K sales points according to the replenishment sequence; When there is a target revenue value greater than a preset threshold among the target revenue values corresponding to each replenishment sequence, determining the candidate replenishment quantities in the replenishment sequence corresponding to the maximum target revenue value as the target replenishment quantities for the first K sales points.

5. The method according to claim 4, characterized in that, the method further includes: After replenishing the top K sales points according to the target replenishment quantity, if the target revenue value within the second preset duration is less than the reference revenue value, then adjust the value of the coverage rate, and return to execute the step of determining the top K sales points from the replenishment priority sequence according to the adjusted coverage rate.

6. The method according to claim 1, wherein, the out-of-stock types further include the normal in-sale type, the current store inventory data of the goods to be replenished at the out-of-stock sales points of the out-of-stock type is equal to zero, the current store inventory data of the goods to be replenished at the sales points of the about-to-be-out-of-stock type is greater than zero and less than or equal to the first threshold, the current store inventory data of the goods to be replenished at the sales points of the normal in-sale type is greater than the first threshold and the historical sales volume is less than the second threshold, and the evaluation values include a first evaluation value, a second evaluation value, and a third evaluation value; The step of inputting the historical sales and inventory data sequences of each sales point into the replenishment prediction models corresponding to different out-of-stock types to obtain the evaluation values of the multiple sales points includes: Inputting the historical sales and inventory data sequences of the goods to be replenished at the out-of-stock sales points of the out-of-stock type into the first binary classification prediction model to obtain the first evaluation value; Inputting the historical sales and inventory data sequences of the goods to be replenished at the sales points of the about-to-be-out-of-stock type into the second binary classification prediction model to obtain the second evaluation value; Inputting the historical sales and inventory data sequences of the goods to be replenished at the sales points of the normal in-sale type into the third prediction model to obtain the third evaluation value, wherein the third prediction model is an average value model.

7. The method according to claim 6, wherein, the training method of the first binary classification prediction model includes the following steps: Obtain a first training sample set, the first training sample set includes the first sales feature data of the sales points where the goods to be replenished are characterized as the out-of-stock type in the historical state within the third preset duration after replenishment, the first sales feature data includes first sales volume data, the first training sample set includes a first positive sample set and a first negative sample set, the first positive sample set includes the first sales feature data of the sales points where the first sales volume data is greater than the third threshold, and the first negative sample set includes the first sales feature data of the sales points where the first sales volume data is less than or equal to the third threshold; Input the first training sample set into the initial first binary classification prediction model, and use the cross-entropy loss function to perform iterative training on the initial first prediction model to obtain the first binary classification prediction model.

8. The method according to claim 6, wherein, the training method of the second binary classification prediction model includes the following steps: Obtain a second training sample set, where the second training sample set includes second sales feature data within a third preset duration after replenishment at a sales point where the in-store inventory data of the product to be replenished in the historical state is greater than zero and less than or equal to the first threshold, and the second sales feature data includes second sales volume data; the second training sample set includes a second positive sample set and a second negative sample set, the second positive sample set includes the second sales feature data of sales points with a difference data greater than a third threshold, and the second negative sample set includes the second sales feature data of sales points with the difference data less than or equal to the third threshold, and the difference data is the difference between the in-store inventory data of the sales point before replenishment and the second sales data within the third preset duration after replenishment; Input the second training sample set into an initial second binary classification prediction model, and use a cross-entropy loss function to iteratively train the initial second prediction model to obtain the second binary classification prediction model.

9. A replenishment device, characterized in that, the device includes: an acquisition module configured to acquire historical sales and inventory data sequences of multiple sales points with different out-of-stock types within a first preset duration; an obtaining module configured to input the historical sales and inventory data sequences of each sales point into replenishment prediction models corresponding to different out-of-stock types to obtain evaluation values of the multiple sales points, where the multiple out-of-stock types at least include an out-of-stock type and an impending out-of-stock type, and the replenishment prediction models corresponding to the out-of-stock type and the impending out-of-stock type are both binary classification models; a first determination module configured to determine a replenishment priority sequence of the multiple sales points according to the evaluation values; a second determination module configured to replenish the multiple sales points according to the replenishment priority sequence.

10. An electronic device, characterized in that, the electronic device includes: a first memory and a first processor; the first memory is used to store program code and transmit the program code to the first processor; the first processor is used to execute the method according to any one of claims 1-8 according to the instructions in the program code.

Images