Commodity sales volume prediction method and device, equipment and storage medium

By building a product cluster and comprehensively considering factors such as historical sales, attributes and promotional activities of similar products, the accuracy of sales forecasts is improved, the problem of inaccurate sales forecasts in the existing technology is solved, and the loss of fresh products is reduced.

CN118982383APending Publication Date: 2024-11-19SHANGHAI 100 METERS NETWORK TECH CO LTD
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411097759.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-12
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

The existing sales forecasting methods are based on a single sequence perspective and cannot capture the substitution and conversion relationship between similar products, resulting in low accuracy of sales forecasts, especially the loss of fresh products.

Method used

By determining the product cluster, consider the historical sales, attribute information and promotional activities of the products to be predicted and other products in the same cluster, and combine the weather and date order situations to build a more comprehensive feature set for sales forecasting.

Benefits of technology

It improves the accuracy of product sales forecasts, reduces the loss rate of fresh products, and optimizes inventory management and procurement decisions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118982383A_ABST
    Figure CN118982383A_ABST
Patent Text Reader

Abstract

The invention provides a commodity sales volume prediction method and device, equipment and a storage medium. The method comprises the steps of determining a commodity cluster to which a to-be-predicted commodity belongs in a selling point; wherein the commodities in the same commodity cluster have a mutual substitution relationship; determining N commodities which are sold on the same day as the to-be-predicted commodity in the selling point from the commodity cluster; determining a first feature set according to the historical sales volume conditions of the N commodities; determining a second feature set according to the attribute information of the N commodities and / or the attribute information of the selling points; inputting the feature set corresponding to the to-be-predicted commodity into the sales prediction model, and obtaining an overall sales prediction value of the commodity cluster at the sales point; the feature set corresponding to the to-be-predicted commodity comprises a first feature set and a second feature set; and splitting the overall sales volume predicted value according to the historical sales volume condition of each commodity in the N commodities, and determining the sales volume predicted value of the to-be-predicted commodity. By adopting the method, the accuracy of a sales volume prediction result can be improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a commodity sales forecasting method, device, equipment and storage medium. Background Art

[0002] The accuracy of sales forecast is crucial for product stocking and procurement, and is also related to the stability of the supply chain distribution system. If the accuracy of product sales forecast is high, it can not only reduce the product defect rate and product maintenance and management costs, but also improve the storage capacity utilization rate of the forward warehouse and avoid unexpected situations such as warehouse explosion.

[0003] Existing sales forecasting methods usually start from a single sequence perspective, and predict the sales of the product in the future based on the historical sales and fluctuation cycle of the product itself. In fact, among the products in the forward warehouse, the sales of some similar products are mutually substitutable and have a certain conversion relationship. However, the existing sales forecasting methods cannot capture this substitution and conversion relationship. Therefore, the accuracy of sales forecasting is low, which is likely to cause a large loss of fresh products. Summary of the invention

[0004] The present application provides a commodity sales forecasting method, apparatus, device and storage medium to improve the accuracy of commodity sales forecasting results.

[0005] In a first aspect, the present application provides a commodity sales forecasting method, the method comprising:

[0006] Determine the commodity cluster to which the commodity to be predicted belongs in the sales point; wherein commodities in the same commodity cluster have a mutual substitution relationship;

[0007] Determine, from the commodity cluster, N commodities sold at the sales outlet on the same day as the commodity to be predicted; the N commodities include the commodity to be predicted, and N is a positive integer;

[0008] Determine a first feature set according to the historical sales of the N commodities;

[0009] Determine a second feature set according to the attribute information of the N commodities and / or the attribute information of the selling point;

[0010] Inputting the feature set corresponding to the commodity to be predicted into a sales prediction model to obtain the overall sales prediction value of the commodity cluster at the sales point; the feature set corresponding to the commodity to be predicted includes the first feature set and the second feature set;

[0011] The overall sales forecast value is split according to the historical sales of each of the N commodities to determine the sales forecast value of the commodity to be forecasted.

[0012] Compared with predicting the sales volume of a commodity based solely on the historical sales volume of the commodity to be predicted, the commodity sales volume prediction method proposed in the present application not only takes into account the historical sales volume of the commodity to be predicted, but also takes into account the historical sales volume of other commodities belonging to the same commodity cluster as the commodity to be predicted, because other commodities belonging to the same commodity cluster as the commodity to be predicted will affect the sales volume of the commodity to be predicted. For example, assuming that a forward warehouse sells fruits such as apples, pears, and peaches, which belong to the same commodity cluster, after the apples in the forward warehouse are sold out, the user may choose pears, peaches, or other fruits; or, the user originally planned to buy apples, but may have purchased other fruits after arriving at the forward warehouse. In other words, the sales volume prediction value of the commodity to be predicted is It is correlated with the sales of other commodities in the same commodity cluster; in addition, the present application also takes into account the impact of the attribute information of the commodities and the attribute information of the selling point on the sales prediction results, wherein the historical sales of commodities in the same commodity cluster corresponds to the first feature set, and the attribute information of the commodities in the same commodity cluster or the attribute information of the selling point corresponds to the second feature set. Therefore, the overall sales forecast value of the commodity cluster determined by the present application according to the first feature set and the second feature set has high accuracy, and the sales forecast value of the commodity to be predicted obtained by splitting the overall sales forecast value is also highly accurate. Guiding the purchase of the commodity to be predicted based on more accurate sales forecast values ​​will also reduce the loss rate of the commodity to be predicted.

[0013] In a possible design, it also includes: determining a third feature set based on the promotional activities set by the sales point for the N commodities; the promotional activities include the number of promotional activities set by the sales point for the N commodities, and / or the discount rate of the promotional activities set by the sales point for the N commodities; the feature set corresponding to the commodity to be predicted also includes the third feature set.

[0014] Generally speaking, the more promotional activities a sales point sets for N commodities, the more sales of N commodities will be; the greater the discount rate of the promotional activities set for N commodities, the more sales of N commodities will be; that is, the sales volume of commodities is related to the number and discount rate of promotional activities set for N commodities by the sales point. Therefore, the third feature set determined according to the promotional activities can also be added to the feature set corresponding to the commodity to be predicted, and input into the sales prediction model together, so as to further improve the accuracy of the sales prediction results.

[0015] In a possible design, it also includes: determining a fourth feature set based on the order situation of the sales point on different days of the week; the order situation includes the order quantity of at least one of the N commodities, and / or the order price of at least one of the N commodities, and / or the business days of the sales point; the feature set corresponding to the commodity to be predicted also includes the fourth feature set.

[0016] Generally speaking, the sales volume of goods on weekends is greater than that on weekdays, that is, there is a difference in the sales volume of N goods on weekends and weekdays. In order to take the impact of this difference on the prediction results into account, the fourth feature set can also be determined according to the order situation of the sales point on different days of the week, and the fourth feature set is added to the feature set corresponding to the product to be predicted, and the fourth feature set is input into the sales prediction model together to further improve the accuracy of the sales prediction results.

[0017] In a possible design, it also includes: determining a fifth feature set based on weather conditions in the area where the geographical location of the sales point belongs; the weather conditions include at least one of precipitation, air quality, weather forecast warning duration, and weather forecast warning level; and the feature set corresponding to the product to be predicted also includes the fifth feature set.

[0018] When the weather conditions are bad, users may go to the sales point another day to purchase items, for example, in heavy rain, severe air pollution, strong wind warning weather, high temperature warning weather, etc. Therefore, the weather conditions in the area where the sales point is located are also related to the sales volume of N products. Therefore, the present application also determines the fifth feature set based on the weather conditions in the area where the geographical location of the sales point is located, and adds the fifth feature set to the feature set corresponding to the product to be predicted, and inputs it into the sales prediction model together to further enhance the accuracy of the feature sales prediction results.

[0019] In one possible design, the historical sales of the N commodities include at least one of the sales corresponding to the set sorting position in the historical sales sorting of each of the N commodities, the sum of the historical sales of the N commodities in a first set dimension, and the sum of the historical daily sales of the N commodities in a second set dimension.

[0020] In a possible design, the attribute information of the N commodities includes a historical average price of each of the N commodities, and / or a penetration rate of M of the N commodities; wherein the M commodities are the M commodities with the highest historical sales volume among the N commodities, and the penetration rate of the M commodities is determined according to the historical sales volume and historical purchase volume of each of the M commodities; M is less than or equal to N;

[0021] The attribute information of the sales point includes the total historical sales amount of the N commodities at the sales point and / or the consumption power of the city where the sales point is located for the N commodities.

[0022] In a possible design, before splitting the overall sales forecast value according to the historical sales of each of the N commodities, the method further includes:

[0023] Obtaining the inventory capacity reserved by the sales point for the commodity cluster;

[0024] If the overall sales forecast value is greater than the inventory capacity, the overall sales forecast value is modified to the inventory capacity.

[0025] In a second aspect, the present application further provides a commodity sales forecasting device, the device comprising: a determination unit and a forecasting unit;

[0026] The determination unit is used to determine the commodity cluster to which the commodity to be predicted belongs in the sales point; wherein the commodities in the same commodity cluster have a mutual substitution relationship; determine N commodities sold on the same day as the commodity to be predicted in the sales point from the commodity cluster; the N commodities include the commodity to be predicted; determine a first feature set based on the historical sales of the N commodities; and determine a second feature set based on the attribute information of the N commodities and / or the attribute information of the sales point;

[0027] The prediction unit is used to input the feature set corresponding to the commodity to be predicted into the sales prediction model to obtain the overall sales prediction value of the commodity cluster at the sales point; the feature set corresponding to the commodity to be predicted includes the first feature set and the second feature set; and the overall sales prediction value is split according to the historical sales of each of the N commodities to determine the sales prediction value of the commodity to be predicted.

[0028] In one possible design, the determination unit is further used to determine a third feature set based on promotional activities set by the selling point for the N commodities; the promotional activities include the number of promotional activities set by the selling point for the N commodities, and / or the discount rate of the promotional activities set by the selling point for the N commodities; the feature set corresponding to the commodity to be predicted also includes the third feature set.

[0029] In a possible design, the determination unit is further used to determine a fourth feature set based on the order status of the selling point on different days of a week; the order status includes the order quantity of at least one of the N commodities, and / or the order price of at least one of the N commodities, and / or the business days of the selling point; the feature set corresponding to the commodity to be predicted also includes the fourth feature set.

[0030] In a possible weather forecast alarm design, the determination unit is also used to determine a fifth feature set based on weather conditions in the area to which the geographical location of the sales point belongs; the weather conditions include at least one of precipitation, air quality, duration, and weather forecast alarm level; and the feature set corresponding to the commodity to be predicted also includes the fifth feature set.

[0031] In one possible design, the historical sales of the N commodities include at least one of the sales corresponding to the set sorting position in the historical sales sorting of each of the N commodities, the sum of the historical sales of the N commodities in a first set dimension, and the sum of the historical daily sales of the N commodities in a second set dimension.

[0032] In a possible design, the attribute information of the N commodities includes a historical average price of each of the N commodities, and / or a penetration rate of M of the N commodities; wherein the M commodities are the M commodities with the highest historical sales volume among the N commodities, and the penetration rate of the M commodities is determined according to the historical sales volume and historical purchase volume of each of the M commodities; M is less than or equal to N;

[0033] The attribute information of the sales point includes the total historical sales amount of the N commodities at the sales point and / or the consumption power of the city where the sales point is located for the N commodities.

[0034] In a possible design, before the overall sales forecast value is split according to the historical sales of each of the N commodities, the forecasting unit is also used to obtain the inventory capacity reserved by the selling point for the commodity cluster; if the overall sales forecast value is greater than the inventory capacity, the overall sales forecast value is modified to the inventory capacity.

[0035] In a third aspect, the present application further provides a commodity sales forecasting device, the device comprising: a processor, and a memory communicatively connected to the processor;

[0036] The memory stores computer-executable instructions;

[0037] The processor executes the computer-executable instructions stored in the memory to implement the method described in the first aspect above.

[0038] In a fourth aspect, the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium includes a program, and when the program is executed on a device, the device executes the method as described in any one of the above-mentioned first aspects.

[0039] In a fifth aspect, the present application further provides a computer program product, wherein the computer program product comprises a computer program, and when the computer program is executed by a processor, the method described in the first aspect is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0041] Figure 1 A flowchart of a method for predicting commodity sales volume provided in an embodiment of the present application;

[0042] Figure 2 A schematic diagram of a feature set provided for an embodiment of the present application;

[0043] Figure 3 Another schematic diagram of the flow chart of the commodity sales forecasting method provided in the embodiment of the present application;

[0044] Figure 4 A schematic diagram of the structure of the commodity sales forecasting device provided in the embodiment of the present application Figure 1 ;

[0045] Figure 5 A schematic diagram of the structure of the commodity sales forecasting device provided in the embodiment of the present application Figure 2 . DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0047] The application scenarios described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. It is known to those skilled in the art that with the emergence of new application scenarios, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems. In the description of the present application, unless otherwise specified, the meaning of "multiple" is two or more.

[0048] As described in the background technology, the sales forecasting method from a single sequence perspective predicts the sales volume of a product in the future based on its own historical sales volume and fluctuation cycle. However, this method does not take into account the product combination and the correlation and substitution relationship between products, and fails to consider the sales transfer problem from the perspective of product clusters. Therefore, the accuracy of sales forecasting results is generally not high, which will lead to greater losses of fresh products.

[0049] For example, taking lean meat, meat with skin, pork belly, and minced meat as examples, from the user's perspective, if the user buys lean meat, he may not buy pork belly, and if he buys pork belly, he may not buy minced meat. In other words, when the predicted sales of alternative products are sufficient, the predicted sales results of the product should be relatively lower. Especially for fresh products, most of them are vegetables, meat, fruits, etc. Most of the fresh products have a short shelf life and cannot maintain a good storage state for a long time. If the sales forecast results of the product are inaccurate, it will directly affect the stocking and procurement of the product, resulting in a large loss of fresh products.

[0050] In order to solve the problem of inaccurate sales forecast results from a single sequence perspective, this application proposes the following Figure 1 The commodity sales forecasting method shown in the figure can be executed by a server, or by a chip or a functional module in the server, which is not limited in this application. The following description is made by taking the server as the execution subject as an example. The method includes:

[0051] Step 100: Determine the commodity cluster to which the commodity to be predicted belongs in the sales point; wherein commodities in the same commodity cluster have a mutual substitution relationship.

[0052] For example, if you want to predict the sales volume of a certain product in a sales point, the product is called the product to be predicted; the sales point can be a forward warehouse or a store, etc. A forward warehouse generally refers to a leased community basement or a small warehouse, which is densely built around the community and is generally within 3 kilometers from the community. The forward warehouse stores a large amount of fresh goods or fast-moving consumer goods, and then the riders are responsible for delivering the goods to consumers' homes, which can meet consumers' needs for convenient and healthy fresh food.

[0053] For example, a product cluster can also be called a Standard Product Unit (SPU). A product cluster is the smallest unit of product information aggregation, and is a set of reusable, easily searchable standardized information that describes the characteristics of a product. In other words, products with the same attribute values ​​and characteristics can be called a product cluster, that is, a product cluster aggregates multiple products with the same attributes. Therefore, each product in the same product cluster is interchangeable. Each product in a product cluster can also be called a Stock Keeping Unit (SKU). In order to distinguish different products in the same product cluster, each product has a different product code to uniquely identify each product.

[0054] For example, lean meat, meat with skin, pork belly, minced meat, etc., they are all pork and have the same attributes, therefore, they belong to the same commodity cluster; apples, pears, peaches, oranges, they are all fruits and have the same attributes, therefore, they also belong to the same commodity cluster.

[0055] Exemplarily, the commodity cluster to which the commodity to be predicted belongs can be determined based on the commodity to be predicted. Specifically, the commodity cluster to which the commodity to be predicted belongs can be determined based on the correspondence between the commodity to be predicted and the commodity cluster. For example, assuming that the correspondence represents the relationship between the number of the commodity to be predicted and the number of the commodity cluster, then the number of the commodity cluster to which the commodity to be predicted belongs can be determined based on the number of the commodity to be predicted, that is, the commodity cluster to which the commodity to be predicted belongs is determined.

[0056] Step 101: Determine N commodities sold at a sales outlet on the same day as the commodity to be predicted from the commodity cluster; the N commodities include the commodity to be predicted, and N is a positive integer.

[0057] Exemplarily, step 100 can be used to determine the commodity cluster to which the commodity to be predicted belongs, and then all the commodities in the commodity cluster can also be determined based on the commodity cluster. However, it is possible that not all the commodities in the commodity cluster are sold at the sales point, and commodities that are not sold will not affect the sales volume of the commodity to be predicted. Commodities that are not sold on the same day as the commodity to be predicted will not affect the sales volume of the commodity to be predicted. Therefore, N commodities sold at the sales point on the same day as the commodity to be predicted should be determined from the commodity cluster, where N is a positive integer and is less than or equal to the number of all the commodities in the commodity cluster.

[0058] Example 1: Assume that there are 100 commodities in the commodity cluster to which the commodity to be predicted belongs, but only 86 of these 100 commodities are sold at the sales point, then N is 86, and these 86 commodities include the commodity to be predicted.

[0059] Example 2: Assume that there are 50 commodities in the commodity cluster to which the commodity to be predicted belongs, and all of these 50 commodities are sold at the sales outlet. Then N is 50, and these 50 commodities include the commodity to be predicted.

[0060] Through this step, the products that do not affect the sales volume of the product to be predicted are removed from the product cluster to which the product to be predicted belongs, thereby reducing the workload in the subsequent determination of the feature set and improving the efficiency of product sales prediction; it also avoids the interference of irrelevant products on the prediction results and improves the accuracy of the prediction results of the product to be predicted.

[0061] Step 102: Determine a first feature set based on historical sales of N products.

[0062] Exemplarily, the historical sales of N commodities include at least one of the sales corresponding to the set ranking position in the historical sales ranking of each of the N commodities, the sum of the historical sales of the N commodities in the first set dimension, and the sum of the historical daily sales of the N commodities in the second set dimension; wherein, the set ranking position can be determined by percentile, for example, a ranking position can be determined by the quantile 10%, or a ranking position can be determined by the quantile 20%...; the sum of the historical sales of N commodities in the first set dimension can be understood as the average of the sum of the maximum historical sales of each of the N commodities, or the average of the sum of the minimum historical sales of each of the N commodities; the sum of the historical daily sales of N commodities in the second set dimension can be understood as the average of the sum of the historical sales of the N commodities on the previous day, or the average of the sum of the historical sales of the N commodities in the previous seven days.

[0063] According to the historical sales of N products, the first feature set can be determined by referring to Table 1:

[0064]

[0065]

[0066] Table 1

[0067] The following is an explanation of the features in Table 1 from top to bottom:

[0068] ①percentile_10: For each of the N commodities, percentile_10 first takes the historical sales of the preset number of days for sorting. For example, taking the first commodity among the N commodities as an example, assuming that the historical sales of the commodity in the first 50 days are sorted from low to high, then the percentile_10 of the commodity is the sales ranked at the 50×10% position, that is, the sales ranked at the 5th position. Assuming that the sales of the first commodity ranked at the 5th position is s1, the corresponding percentile_10 can be determined in the same way for the remaining 2nd to N commodities among the N commodities. Assuming that the corresponding percentile_10 of the 2nd to N commodities is S2~S N, then for these N commodities, percentile_10 is the average value of the sum of percentile_10 corresponding to each commodity in the N commodities, that is,

[0069] ②percentile_25: Similarly, assuming that the historical sales of N products in the previous 50 days are still sorted from low to high, taking the first product among the N products as an example, the percentile_25 of the product is the sales ranked at the 50×25% position. Since 50×25% is 12.5, the non-integer sorting position is usually rounded off, that is, the percentile_25 of the product is the sales ranked at the 13th position (other methods can also be used to process the non-integer sorting position, which is not limited in this application). Assuming that the sales of the first product ranked at the 13th position is M1, the corresponding percentile_25 of the remaining 2nd to N products among the N products can be determined in the same way. Assuming that the percentile_25 corresponding to the 2nd to N products is M2~M N , then for these N commodities, percentile_25 is the average value of the sum of percentile_25 corresponding to each commodity in the N commodities, that is,

[0070] ③The calculation method of percentile_50 is the same as that of percentile_10 and percentile_25, so it will not be repeated here.

[0071] ④The calculation method of percentile_75 is the same as that of percentile_10 and percentile_25, so it will not be repeated here.

[0072] ⑤The calculation method of percentile_95 is the same as that of percentile_10 and percentile_25, so it will not be repeated here.

[0073] ⑥max_sales: refers to the average of the maximum historical sales of each product in N products. For example, suppose the maximum sales of the first product in the N products in the past 14 days is K1, and the maximum sales of the second product in the N products in the past 14 days is K2... The maximum historical sales of the Nth product is K N , then max_sales is K1~K N The average value of the sum, that is

[0074] ⑦min_sales: refers to the average of the minimum sales of each product in N products. For example, suppose the minimum sales of the first product in N products in the past 14 days is P1, and the minimum sales of the second product in N products in the past 14 days is P2... The minimum sales of the Nth product in the past 14 days is P N , then min_sales is P1~P N The average value of the sum, that is

[0075] ⑧spu_max_sales: refers to summing up the historical daily sales of each product in N products, and taking the maximum value of the sum of historical daily sales. For example, suppose the sales of N products in the past 14 days are counted, and the sales of N products in each of the 14 days are summed up to obtain 14 summation results, which are recorded as Sum1~Sum 14 , Sum1~Sum 14 The maximum value in is spu_max_sales.

[0076] ⑨spu_min_sales: refers to summing up the historical daily sales of each product among N products and taking the minimum value of the sum of historical daily sales.

[0077] ⑩std_sales: Use the standard deviation calculation formula to first calculate the sales standard deviation of each of the N products, and then sum and average the sales standard deviations of each product. For example, first calculate the historical standard deviation of the first product based on the daily sales of the first product in the previous 14 days and the average sales of the product in the previous 14 days. Use the same calculation method to calculate the sales standard deviation of the second to N products, and then sum and average the sales standard deviations corresponding to the first to N products.

[0078] avg_sales: refers to the average of the sum of sales of N products on the previous day.

[0079] avg_7d: refers to the average of the sales volume of N products in the previous seven days.

[0080] avg_7d_ignore: If the sales outlet has set up a promotion for N products on a certain day in the previous seven days, then the sales volume of the N products on that day will be subtracted from the sum of the sales volume of the previous seven days and then the average will be calculated. For example, assuming that the sum of the sales volume of N products in the previous seven days is 1000, and the sales outlet has set up a promotion for these N products last Wednesday, and the sales volume of these N products last Wednesday was 200, then avg_7d_ignore will subtract 200 from 1000 and then calculate the average.

[0081] Step 103: Determine a second feature set based on the attribute information of the N commodities and / or the attribute information of the selling point.

[0082] Exemplarily, the attribute information of N commodities may include the historical average price of each of the N commodities. For example, taking the past ten days as an example, first calculate the average price of the first commodity in the N commodities in the past ten days, then calculate the average price of the second commodity in the past ten days... calculate the average price of the Nth commodity in the past ten days, and then sum and average the corresponding average prices of the N commodities.

[0083] Exemplarily, the attribute information of N commodities may also include the penetration rate of M commodities among the N commodities; wherein, the M commodities are the M commodities with the highest historical sales volume among the N commodities, and the penetration rate of the M commodities is determined based on the historical sales volume and historical purchase volume of each commodity among the M commodities, and M is less than or equal to N. For example, if M is 20, then the penetration rate of the M commodities among the N commodities is the penetration rate of the top 20 commodities in historical sales volume among the N commodities. The penetration rate of the commodity may be determined by historical sales volume / historical purchase volume. For example, the historical sales volume of a commodity among the M commodities is 180, and the historical purchase volume is 200. If the above-mentioned penetration rate calculation formula is used, the penetration rate of the commodity is 180 / 200, which is 0.9.

[0084] Exemplarily, the attribute information of the sales point may include the historical total sales amount of N commodities at the sales point, and the historical total sales amount may also be represented by Gross Merchandise Volume (GMV).

[0085] Exemplarily, the attribute information of the sales point may also include the consumption power of the city where the sales point is located for N commodities. The ranking of the historical total sales of the N commodities in the city where the sales point is located may be used as the consumption power. For example, assuming that the city where the sales point is located is Shanghai, the historical total sales of N commodities in Shanghai is L1, and the historical total sales of the N commodities in Beijing is L2... Similarly, the historical total sales of the N commodities in other cities may also be obtained. Assuming that L1 ranks fourth in the historical total sales of the N commodities in each city (assuming sorting from high to low), the ranking may be used as the consumption power of the city where the sales point is located for the N commodities.

[0086] Exemplarily, the attribute information of the sales point may also include the historical total sales amount of N commodities at the sales point last weekend.

[0087] The second feature set determined based on the attribute information of N commodities and the attribute information of the selling point can be referred to in Table 2:

[0088] Feature Name type Field meaning avg_origin_price float Original average price store_avg_top20_permeability float TopM penetration store_gmv float GMV of forward warehouse store_rank_by_gmv_city int City GMV ranking store_gmv_weekend float Weekend forward warehouse gmv

[0089] Table 2

[0090] Among them, avg_origin_price in Table 2 is the historical average price of each of the N commodities, store_avg_top20_permeability is the permeability of M commodities among the N commodities; store_gmv is the historical total sales of the N commodities at the sales point; store_rank_by_gmv_city is the consumption power of the city where the sales point is located for the N commodities; store_gmv_weekend is the historical total sales of the N commodities last weekend.

[0091] Step 104: Input the feature set corresponding to the commodity to be predicted into the sales prediction model to obtain the overall sales prediction value of the commodity cluster at the sales point; the feature set corresponding to the commodity to be predicted includes a first feature set and a second feature set.

[0092] Exemplarily, after obtaining the first feature set and the second feature set through step 103, the feature set including the first feature set and the second feature set can be input into the sales prediction model. The output of the sales prediction model is the overall sales prediction value of the commodity cluster at the sales point. The so-called overall sales prediction value can be understood as the sales prediction value corresponding to N commodities as a whole.

[0093] Exemplarily, in order to improve the comprehensiveness of the feature data contained in the feature set corresponding to the commodity to be predicted and improve the accuracy of the prediction result, the present application also provides a more preferred implementation method for determining the feature set corresponding to the commodity to be predicted.

[0094] Specifically, the present application may also determine the third feature set according to the promotional activities set by the sales point for N commodities, wherein the promotional activities may include the number of promotional activities set by the sales point for N commodities, and may also include the discount rate of the promotional activities set by the sales point for N commodities. After determining the third feature set according to the promotional activities, the third feature set is added to the feature set corresponding to the commodity to be predicted, so that the feature set corresponding to the commodity to be predicted also includes the third feature set. The third feature set can refer to Table 3:

[0095] Feature Name type Field meaning activity_cnt int Number of activities weighted_avg_discount_activity float TopM penetration

[0096] Table 3

[0097] Among them, activity_cnt in Table 3 is the number of promotional activities set by the sales point for N commodities, and is the discount rate of the promotional activities set by the sales point for N commodities.

[0098] Generally speaking, the more promotional activities a sales point sets for N commodities, the more sales of N commodities will be; the greater the discount rate of the promotional activities set for N commodities, the more sales of N commodities will be; that is, the sales volume of commodities is related to the number and discount rate of promotional activities set for N commodities by the sales point. Therefore, the third feature set determined according to the promotional activities is added to the feature set corresponding to the commodity to be predicted. The sales prediction model can make predictions based on a more comprehensive feature set, thereby improving the accuracy of the output of the sales prediction model, that is, improving the accuracy of the sales prediction results.

[0099] Exemplarily, the present application may also determine a fourth feature set based on the order status of the sales point on different days of the week, wherein the order status may include the number of orders with at least one of the N items, and the number of orders may be represented by the average number of orders over historical weekends. For example, assuming the historical two weeks as an example, if the number of orders for at least one of the N items at the sales point in the previous weekend was 100; if the number of orders for at least one of the N items at the sales point in the previous weekend was 150, then the number of orders for at least one of the N items is (100+150) / 2, which is 225.

[0100] Exemplarily, the order status may also include the order price of at least one of the N commodities, and the order price may be composed of the weekend average order value and the seven-day average order value. First, the weekend average order value is explained. Assuming that the sales point has 40 orders for at least one of the N commodities in two historical weekends, and the order amount is 2,000 yuan in total, then the weekend average order value is order amount / order quantity = 2,000 / 40 = 50 yuan; then the seven-day average order value is explained. The seven-day average order value and the weekend average order value are similar in meaning, except that the statistical time period is different. For example, assuming that the sales point has 60 orders for at least one of the N commodities in one historical week, and the order amount is 2,400 yuan in total, then the weekend average order value is order amount / order quantity = 2,400 / 60 = 40 yuan.

[0101] Exemplarily, the order status may also include the business days of the sales point. Generally speaking, the longer the business hours of a sales point, the better the sales of the goods. In other words, the sales of N goods are positively correlated with the business days of the sales point. In addition, the order status may also include the number of days in a week that the sales point sells at least one of the N goods. For example, assuming that N is 3, and the N goods are apples, pears, and peaches, assuming that the sales point sells apples on Monday, peaches on Tuesday, and pears on Saturday, then the number of days in a week that the sales point sells at least one of the N goods is 3 days; assuming that the sales point sells apples on Monday, Tuesday, and Wednesday, and sells pears and peaches from Thursday to Sunday, then the number of days in a week that the sales point sells at least one of the N goods is 7 days.

[0102] Further exemplary, after the fourth feature set is determined according to the order situation, the fourth feature set is added to the feature set corresponding to the product to be predicted, so that the feature set corresponding to the product to be predicted also includes the fourth feature set. The fourth feature set can refer to Table 4:

[0103]

[0104]

[0105] Table 4

[0106] Among them, store_user_arpa_weekend and store_user_arpa_7d are both the order price of at least one of the N commodities; store_order_amt_weekend is the order quantity of at least one of the N commodities; open_days_amt is the number of days the sales point is open; dayofweek is the number of days in a week when the sales point sells at least one of the N commodities.

[0107] In addition, the sales volume of goods is also related to the weather conditions in the area where the geographical location of the sales point is located. Therefore, this application also determines the fifth feature set based on the weather conditions in the area where the geographical location of the sales point is located; wherein the weather conditions include at least one of precipitation, air quality, weather forecast warning duration, and weather forecast warning level.

[0108] Specifically, the precipitation can be determined based on the average precipitation of the historical preset days in the area where the geographical location of the sales point belongs, for example, based on the average precipitation of the previous five days; the air quality is used to indicate the quality of the air quality in the area where the geographical location of the sales point belongs, and can be expressed by air quality levels, for example, when the air quality level is 1, it means the best air quality. When the air quality level is 10, the air quality is the worst; or when the air quality level is 10, the air quality is the best; when the air quality level is 1, the air quality is the worst; this application does not limit this.

[0109] The weather forecast alarm duration is the sum of the durations of various weather forecast alarms in a day or the proportion of the sum of the durations of various weather forecast alarms in a day. For example, assuming that there are both strong wind forecasts and heavy rain warnings in a day, where the strong wind warning duration is 2 hours and the heavy rain warning duration is 3 hours, if the weather forecast alarm duration is the sum of the durations of various weather forecast alarms in a day, then the weather forecast alarm duration for that day is 5 hours; if the weather forecast alarm duration is the proportion of the sum of the durations of various weather forecast alarms in a day, then the weather forecast alarm duration for that day = 5 hours / 24 hours = 21%.

[0110] The weather forecast alert level is used to indicate the intensity of the weather forecast alert. The alert levels for each day in the historical preset days can be summed up and then the average value can be calculated. For example, taking the strong wind alert as an example, a weather forecast alert level of 10 can indicate the strongest wind force. Taking the heavy rain alert as an example, a weather forecast alert level of 10 can indicate the heaviest rainfall. Assuming the historical thirty days as an example, first count the weather forecast alerts and the corresponding alert levels for each day in the historical thirty days, then sum up all the alert levels, and finally calculate the average value based on the summed result.

[0111] After the fifth feature set is determined according to the weather conditions, the fifth feature set is added to the feature set corresponding to the commodity to be predicted. Thus, the feature set corresponding to the commodity to be predicted also includes the fifth feature set. The fifth feature set can be referred to in Table 5:

[0112]

[0113] Table 5

[0114] Among them, precipitation is the precipitation in the weather conditions; aqi is the air quality in the weather conditions; alert_ratio is the weather forecast warning duration; aqi_div_5_days can be used to indicate the change in air quality in 5 days, for example, the air quality level on July 1 is 3, and the air quality level on July 6 is 8, then aqi_div_5_days is 5; wd_garde_max can be used to indicate the level of the degree of temperature drop within 5 days, for example, if the temperature drops by less than 5 degrees, the temperature drop level is 1; if the temperature drops between 5 and 10 degrees, the temperature drop level is 2; if the temperature drops between 10 and 15 degrees, the temperature drop level is 3...; alert_grade is the weather forecast warning level.

[0115] So far, the feature set corresponding to the product to be predicted includes the first feature set, the second feature set, the third feature set, the fourth feature set and the fifth feature set, wherein the first feature set can also be called lag feature or hysteresis feature, the second feature set can also be called attribute feature, the third feature set can also be called activity feature, the fourth feature set can also be called time feature, and the fifth feature set can also be called weather feature. Therefore, the feature set of the product to be predicted can be expressed as Figure 2 Then, the feature set corresponding to the product to be predicted, that is, the first feature set to the fifth feature set, is input into the sales prediction model, and the overall sales prediction value of the product cluster at the sales point can also be obtained.

[0116] Step 105: split the overall sales forecast value according to the historical sales of each of the N commodities, and determine the sales forecast value of the commodity to be forecasted.

[0117] Exemplarily, through step 104, the overall sales forecast value of the commodity cluster at the sales point can be obtained, and the overall sales forecast value is split according to the historical sales of each of the N commodities included in the commodity cluster, so as to obtain the sales forecast value of each of the N commodities, that is, the sales forecast value of the commodity to be predicted is obtained.

[0118] For example, assuming that the overall sales forecast value is 1000, and the historical sales of the product to be forecast accounts for 1 / 50 of the sum of the historical sales of N products, then the sales forecast value of the product to be forecast is 1000×1 / 50=20.

[0119] Further exemplarily, the selling point has planned corresponding inventory capacity for each commodity cluster. For example, the inventory capacity reserved for commodity cluster A at the selling point is 1000, and the inventory capacity reserved for commodity cluster B at the selling point is 800.....then the situation may occur: the overall sales forecast value of the commodity cluster obtained in step 104 is greater than the inventory capacity reserved for the commodity cluster by the selling point. In practical applications, the above situation should be avoided. Therefore, before the selling point splits the overall sales forecast value according to the historical sales of each of the N commodities, the inventory capacity reserved for the commodity cluster by the selling point is first obtained; if the overall sales forecast value is greater than the inventory capacity, the overall sales forecast value is modified to the inventory capacity.

[0120] For example, assuming that the overall sales forecast value of commodity cluster A is 1000, and the inventory capacity reserved for commodity cluster A at the sales point is 900, then the overall sales forecast value of commodity cluster A should be modified to 900, and then 900 should be split according to the historical sales of each of the N commodities to finally determine the sales forecast value of the commodity to be predicted.

[0121] The process of the product sales forecasting method of this application can also be referred to Figure 3 , Figure 3 The detailed process of merging products in the above-mentioned steps 100 and 101 can be referred to, and will not be repeated here; Figure 3 The detailed process of feature engineering in can refer to the above steps 102 and 103. Figure 3 The detailed process of the overall prediction in can refer to the above step 104. Figure 3 For the detailed process of commodity splitting, please refer to the above step 105.

[0122] Figure 4 and Figure 5 A schematic diagram of the structure of a possible commodity sales forecasting device provided in an embodiment of the present application. These commodity sales forecasting devices can be used to implement the functions of the server in the above method embodiment, and thus can also achieve the beneficial effects possessed by the above method embodiment.

[0123] like Figure 4 As shown, the commodity sales forecasting device 400 includes a determination unit 410 and a forecasting unit 420. The commodity sales forecasting device 400 is used to implement the above Figure 1 The functions of the server in the method embodiment are shown.

[0124] When the commodity sales forecasting device 400 is used to implement Figure 1 The functions of the server in the method embodiment shown are:

[0125] The determination unit 410 is used to determine the commodity cluster to which the commodity to be predicted belongs in the selling point; wherein the commodities in the same commodity cluster have a mutual substitution relationship; determine N commodities sold on the same day as the commodity to be predicted in the selling point from the commodity cluster; the N commodities include the commodity to be predicted; determine a first feature set based on the historical sales of the N commodities; and determine a second feature set based on the attribute information of the N commodities and / or the attribute information of the selling point;

[0126] The prediction unit 420 is used to input the feature set corresponding to the commodity to be predicted into the sales prediction model to obtain the overall sales prediction value of the commodity cluster at the sales point; the feature set corresponding to the commodity to be predicted includes the first feature set and the second feature set; and the overall sales prediction value is split according to the historical sales of each of the N commodities to determine the sales prediction value of the commodity to be predicted.

[0127] In one possible design, the determination unit 410 is further used to determine a third feature set based on the promotional activities set by the selling point for the N commodities; the promotional activities include the number of promotional activities set by the selling point for the N commodities, and / or the discount rate of the promotional activities set by the selling point for the N commodities; the feature set corresponding to the commodity to be predicted also includes the third feature set.

[0128] In a possible design, the determination unit 410 is further used to determine a fourth feature set based on the order status of the selling point on different days of a week; the order status includes the order quantity of at least one of the N commodities, and / or the order price of at least one of the N commodities, and / or the business days of the selling point; the feature set corresponding to the commodity to be predicted also includes the fourth feature set.

[0129] In a possible weather forecast alarm design, the determination unit 410 is also used to determine a fifth feature set based on the weather conditions in the area to which the geographical location of the sales point belongs; the weather conditions include at least one of precipitation, air quality, duration, and weather forecast alarm level; and the feature set corresponding to the commodity to be predicted also includes the fifth feature set.

[0130] In one possible design, the historical sales of the N commodities include at least one of the sales corresponding to the set sorting position in the historical sales sorting of each of the N commodities, the sum of the historical sales of the N commodities in a first set dimension, and the sum of the historical daily sales of the N commodities in a second set dimension.

[0131] In a possible design, the attribute information of the N commodities includes a historical average price of each of the N commodities, and / or a penetration rate of M of the N commodities; wherein the M commodities are the M commodities with the highest historical sales volume among the N commodities, and the penetration rate of the M commodities is determined according to the historical sales volume and historical purchase volume of each of the M commodities; M is less than or equal to N;

[0132] The attribute information of the sales point includes the total historical sales amount of the N commodities at the sales point and / or the consumption power of the city where the sales point is located for the N commodities.

[0133] In a possible design, before the overall sales forecast value is split according to the historical sales of each of the N commodities, the forecasting unit 420 is also used to obtain the inventory capacity reserved by the selling point for the commodity cluster; if the overall sales forecast value is greater than the inventory capacity, the overall sales forecast value is modified to the inventory capacity.

[0134] For a more detailed description of the above-mentioned determination unit 410 and prediction unit 420, please refer directly to Figure 1 The relevant description in the method embodiment shown is directly obtained and will not be repeated here.

[0135] like Figure 5 As shown, the commodity sales forecasting device 500 includes a processor 510 and an interface circuit 520. The processor 510 and the interface circuit 520 are coupled to each other. It can be understood that the interface circuit 520 can be a transceiver or an input-output interface. Optionally, the commodity sales forecasting device 500 can also include a memory 530 for storing instructions executed by the processor 510 or storing input data required by the processor 510 to execute instructions or storing data generated after the processor 510 executes instructions.

[0136] When the commodity sales forecasting device 500 is used to implement Figure 1 When the method is shown, the processor 510 is used to implement the function of the prediction unit 420, and the interface circuit 520 is used to implement the function of the determination unit 410.

[0137] The division of units in the embodiments of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional unit in each embodiment of the present application may be integrated into a processor, or may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0138] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0139] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A commodity sales forecasting method, characterized in that: The method includes: Determine the commodity cluster to which the commodity to be predicted belongs in the sales point; wherein commodities in the same commodity cluster have a mutual substitution relationship; Determine, from the commodity cluster, N commodities sold at the sales outlet on the same day as the commodity to be predicted; the N commodities include the commodity to be predicted, and N is a positive integer; Determine a first feature set according to the historical sales of the N commodities; Determine a second feature set according to the attribute information of the N commodities and / or the attribute information of the selling point; Inputting the feature set corresponding to the commodity to be predicted into a sales prediction model to obtain the overall sales prediction value of the commodity cluster at the sales point; the feature set corresponding to the commodity to be predicted includes the first feature set and the second feature set; The overall sales forecast value is split according to the historical sales of each of the N commodities to determine the sales forecast value of the commodity to be forecasted.

2. The method according to claim 1, characterized in that Also includes: A third feature set is determined based on the promotional activities set by the selling point for the N commodities; the promotional activities include the number of promotional activities set by the selling point for the N commodities, and / or the discount rate of the promotional activities set by the selling point for the N commodities; the feature set corresponding to the commodity to be predicted also includes the third feature set.

3. The method according to claim 1 or 2, characterized in that Also includes: Determining a fourth feature set according to the order status of the sales point on different days of the week; The order status includes the order quantity of at least one of the N commodities, and / or the order price of at least one of the N commodities, and / or the business days of the sales point; the feature set corresponding to the commodity to be predicted also includes the fourth feature set.

4. The method according to claim 3, characterized in that Also includes: A fifth feature set is determined based on the weather conditions in the area where the geographical location of the sales point belongs; the weather conditions include at least one of precipitation, air quality, weather forecast warning duration, and weather forecast warning level; and the feature set corresponding to the product to be predicted also includes the fifth feature set.

5. The method according to claim 1, characterized in that The historical sales of the N commodities include at least one of the sales corresponding to the set ranking position in the historical sales ranking of each of the N commodities, the sum of the historical sales of the N commodities in a first set dimension, and the sum of the historical single-day sales of the N commodities in a second set dimension.

6. The method according to claim 1, characterized in that The attribute information of the N commodities includes a historical average price of each of the N commodities, and / or a penetration rate of M of the N commodities; wherein the M commodities are the M commodities with the highest historical sales volume among the N commodities, and the penetration rate of the M commodities is determined according to the historical sales volume and historical purchase volume of each of the M commodities; M is less than or equal to N; The attribute information of the sales point includes the total historical sales amount of the N commodities at the sales point and / or the consumption power of the city where the sales point is located for the N commodities.

7. The method according to claim 1, characterized in that Before splitting the overall sales forecast value according to the historical sales of each of the N commodities, the method further includes: Obtaining the inventory capacity reserved by the sales point for the commodity cluster; If the overall sales forecast value is greater than the inventory capacity, the overall sales forecast value is modified to the inventory capacity.

8. A commodity sales forecasting device, characterized in that: The device comprises: a determination unit and a prediction unit; The determination unit is used to determine the commodity cluster to which the commodity to be predicted belongs in the sales point; wherein the commodities in the same commodity cluster have a mutual substitution relationship; determine N commodities sold on the same day as the commodity to be predicted in the sales point from the commodity cluster; the N commodities include the commodity to be predicted, and N is a positive integer; determine a first feature set based on the historical sales of the N commodities; and determine a second feature set based on the attribute information of the N commodities and / or the attribute information of the sales point; The prediction unit is used to input the feature set corresponding to the commodity to be predicted into the sales prediction model to obtain the overall sales prediction value of the commodity cluster at the sales point; the feature set corresponding to the commodity to be predicted includes the first feature set and the second feature set; and the overall sales prediction value is split according to the historical sales of each of the N commodities to determine the sales prediction value of the commodity to be predicted.

9. A commodity sales forecasting device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 7 when executed by a processor.

11. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 7 when being executed by a processor.

Citation Information

Cited By

  • Intelligent supply chain management optimization method and system based on big data

    CN120851299A

  • Big data-based intelligent supply chain management optimization method and system

    CN120851299B