Sales forecasting method, device and computer equipment

Abstract

The present application provides a sales forecasting method, device and computer equipment, the method comprising obtaining target sales feature information associated with a preset date interval to be predicted from sales feature information of a sales point; inputting the target sales feature information into a pre-trained sales forecasting model, obtaining the average sales of the preset date interval and the daily sales corresponding to each date included in the preset date interval through the sales forecasting model; based on the average sales and daily sales, determining the target daily sales of each date included in the preset date interval at the sales point. The present application can improve the sales forecast accuracy of each date.

Description

Technical Field

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

[0002] With the development of machine learning technology, machine learning is widely used in product sales prediction to enable merchants to better provide enough products for sales. Especially for holidays, it is necessary to ensure that there are enough products for sales.

[0003] In the sales forecasting scenario, we mainly refer to historical sales data, use statistical inference, machine learning models, deep learning models and other solutions to predict future sales, and ultimately provide stocking guidance at the business level.

[0004] However, in the related art, there is at least a problem that the sales forecast results are not accurate enough. Summary of the invention

[0005] Based on this, the purpose of this application is to solve at least one of the above-mentioned technical defects, especially the technical defect of low sales forecast accuracy in the prior art. This application provides a sales forecast method, device and computer equipment.

[0006] In a first aspect, the present application provides a sales volume forecasting method, the method comprising:

[0007] Obtaining target sales characteristic information associated with a preset date interval to be predicted from the sales characteristic information of the sales point;

[0008] Input the target sales feature information into the pre-trained sales forecasting model, and obtain the average sales in the preset date range and the daily sales corresponding to each date in the preset date range through the sales forecasting model;

[0009] Based on the average sales volume and the daily sales volume, the target daily sales volume of the sales point on each date included in the preset date range is determined.

[0010] In one embodiment, obtaining target sales characteristic information associated with a preset date interval to be predicted from sales characteristic information of a sales point includes:

[0011] Obtaining current sales characteristic information of the year in which the preset date interval falls, and historical sales characteristic information corresponding to the preset date interval; wherein the historical sales characteristic information is sales characteristic information of the year in which the preset date interval falls;

[0012] According to the current sales characteristic information and the historical sales characteristic information, the target sales characteristic information is obtained.

[0013] In one embodiment, obtaining current sales characteristic information of the year in which the preset date interval falls and historical sales characteristic information corresponding to the preset date interval includes:

[0014] Get the current sales data and current sales change data for the year of the preset date range and the date before the preset date range;

[0015] Get the historical sales data and historical sales change data corresponding to the preset date range;

[0016] According to the current sales data and the current sales change data, the current sales characteristic information is obtained;

[0017] Based on the historical sales data and historical sales change data, historical sales characteristic information is obtained.

[0018] In one embodiment, obtaining target sales characteristic information associated with a preset date interval to be predicted from sales characteristic information of a sales point includes:

[0019] From the sales feature information of the sales point, obtaining initial sales feature information associated with the preset date interval to be predicted;

[0020] The initial sales feature information is discretely detected and the discrete data is processed to obtain the target sales feature information.

[0021] In one embodiment, the method further comprises:

[0022] Get the last predicted sales volume of the last predicted date interval corresponding to the preset date interval, and the actual sales volume of the last predicted date interval;

[0023] Using the previous predicted sales volume and actual sales volume, the sales volume forecasting model is updated;

[0024] Through the sales forecast model, the average sales volume in the preset date range and the daily sales volume corresponding to each date in the preset date range are obtained, including:

[0025] Through the updated sales forecasting model, the average sales volume in the preset date range and the daily sales volume corresponding to each date in the preset date range are obtained.

[0026] In one embodiment, the target sales characteristic information includes sales characteristics of each product category; and the method further includes:

[0027] Through the sales forecast model, obtain the predicted sales of each product category in the preset date range;

[0028] Get the sales volume of categories in non-preset date intervals other than the preset date interval;

[0029] Based on the category forecast sales and category sales, determine the predicted sales change for each product category.

[0030] In one embodiment, based on the average sales volume and the daily sales volume, determining the target daily sales volume of the sales point on each date included in the preset date interval includes:

[0031] Get the degree of difference between daily sales and average sales for each date;

[0032] Based on the degree of difference of each date, the target daily sales volume of the sales point on each date included in the preset date range is determined.

[0033] In one embodiment, the training method of the sales forecasting model includes:

[0034] From the sales feature information of the sales point, obtain the historical sales feature information associated with the preset date interval to be predicted, and the actual historical daily sales of each date included in the preset date interval to be predicted;

[0035] Input the historical sales feature information into the sales forecasting model to be trained, and obtain the predicted historical average sales for a preset date range and the predicted historical daily sales corresponding to each date in the preset date range through the sales forecasting model;

[0036] Determine the predicted historical daily sales volume of the sales point in a preset date range based on the predicted historical average sales volume and the historical predicted target daily sales volume;

[0037] The sales forecasting model is trained according to the historical forecast target daily sales and the actual historical daily sales to obtain a trained sales forecasting model.

[0038] In a second aspect, the present application provides a sales forecasting device, the device comprising:

[0039] A feature acquisition module, used to acquire target sales feature information associated with a preset date interval to be predicted from sales feature information of a sales point;

[0040] A prediction module is used to input the target sales feature information into a pre-trained sales prediction model, and obtain the average sales in a preset date range and the daily sales corresponding to each date in the preset date range through the sales prediction model;

[0041] The sales volume determination module is used to determine the target daily sales volume of the sales point on each date included in the preset date range based on the average sales volume and the daily sales volume.

[0042] In a third aspect, the present application provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0043] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0044] The sales forecasting method, device and computer equipment provided by the present application can obtain the target sales feature information associated with the preset date interval to be predicted, and predict through the pre-trained sales forecasting model, so as to obtain the average sales of the preset date interval and the daily sales corresponding to each date in the preset date interval, and by using the predicted average sales and the predicted daily sales of each date, it is possible to effectively and accurately determine the target daily sales of each date in the preset date interval. In this way, compared with the traditional technology, the present application predicts the average sales and daily sales through the sales forecasting model, further determines the average sales and daily sales, and determines the target sales predicted for each date, thereby improving the accuracy of the daily sales forecast. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art 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 labor.

[0046] Figure 1 Schematic diagram of the sales forecasting method provided in the embodiment of the present application Figure 1 ;

[0047] Figure 2 A schematic diagram of the flow of sales forecast model training provided in an embodiment of the present application;

[0048] Figure 3 Schematic diagram of the sales forecasting method provided in the embodiment of the present application Figure 2 ;

[0049] Figure 4 Schematic diagram of the sales forecasting method provided in the embodiment of the present application Figure 3 ;

[0050] Figure 5 Schematic diagram of the sales forecasting method provided in the embodiment of the present application Figure 4 ;

[0051] Figure 6 A schematic diagram of the framework of the sales forecasting method provided in the embodiment of the present application;

[0052] Figure 7 A schematic diagram of the structure of a sales forecasting device provided in an embodiment of the present application;

[0053] Figure 8 A schematic diagram of the internal structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0055] With the development of machine learning technology, machine learning is widely used in product sales prediction to enable merchants to better provide enough products for sales. Especially for holidays, it is necessary to ensure that there are enough products for sales.

[0056] In the sales forecasting scenario, we mainly refer to historical sales data, use statistical inference, machine learning models, deep learning models and other solutions to predict future sales, and ultimately provide stocking guidance at the business level.

[0057] MA (Moving Average) is a common sales forecasting model and a benchmark model in many sales forecasting scenarios. Since sales fluctuate greatly during holidays, and the moving average model can only be predicted based on historical sales, it is difficult to simulate daily sales trends during holidays. Some models that can simulate daily sales trends during holidays (such as LGBM, Transformer, etc., which can apply the same holiday data of the previous year) are not adequate for the trend of individual products. Therefore, the relevant technologies have at least the following technical problems:

[0058] It is difficult to fit the sales trend during holidays, and the accuracy of the prediction results is not high; it is difficult to process training data, especially statistical machine learning models suitable for data mining need to build a large amount of feature data; model training takes a long time, especially complex models (deep learning models) have many parameters and therefore require a lot of training time; international holiday prediction is difficult, because different countries and holidays have corresponding special data processing, simple simulation performance is poor and complex models need to be trained according to different data parameters;

[0059] That is, in the related technology, the main problem is that the sales forecast results are not accurate enough.

[0060] Based on this, the present application provides a sales forecasting method, device and computer equipment. The present application predicts the average sales and daily sales through a sales forecasting model, further determines the average sales and daily sales, and determines the target sales predicted for each date, thereby improving the accuracy of daily sales forecasts.

[0061] In an exemplary embodiment, Figure 1 Schematic diagram of the sales forecasting method provided in the embodiment of the present application Figure 1 ,like Figure 1 As shown, a sales forecasting method is provided, and the method is applied to a server as an example for explanation. It can be understood that the method can also be applied to a terminal, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes S101 to S103. Among them:

[0062] S101. Obtain target sales characteristic information associated with a preset date range to be predicted from sales characteristic information of a sales point.

[0063] Among them, the sales point refers to the point where the goods are sold. The sales situation of each sales point is different, so sales forecasts can be made for the sales points. Sales feature information can refer to features related to sales. For example, sales feature information can include sales data, sales change data, sales trend indicators and other information. The preset date interval can refer to the sales date interval to be predicted; as an example, the preset date interval can be a holiday interval, that is, the dates when the holidays last; in addition, in addition to the specified holiday dates, the preset date interval can also include the dates of the days before and after the holidays.

[0064] The target sales characteristic information refers to the sales characteristic information related to the preset date range. For example, the target sales characteristic information can be the sales characteristic information of the dates before the preset date range, or it can be the sales trend characteristics before the preset date range, the sales information of the preset date range in the previous year, etc.

[0065] For example, when the server needs to predict the sales volume of a certain sales point, it can obtain the sales volume characteristic information of the sales point, for example, it can obtain the sales volume characteristic information of the sales point from the sales database. It can be understood that the sales volume characteristic information here can include all the sales volume characteristic information of the sales point, for example, the sales volume characteristic information of this year and the sales volume characteristic information of previous years.

[0066] The server can determine the sales characteristic information associated with the preset date interval to be predicted from the sales characteristic information of the sales point as the target sales characteristic information. For example, the server can determine the sales data and sales change data before the preset date interval in the year in which the preset date interval is located. The sales change data described here can also be a sales trend factor. The trend factor refers to the description of the sales trend of stores and subcategories before the festival this year. The server can also determine the sales data and sales change data of the preset date interval in the previous year.

[0067] Optionally, the server may use the sales data and sales change data of the year in which the preset date interval falls and before the preset date interval, and the sales data and sales change data of the preset date interval in the previous year as the target sales feature information.

[0068] In practical applications, the above method can effectively determine the target sales feature information associated with the preset date range to be predicted, so that sales forecasting can be performed using fewer data features, thereby improving sales forecasting efficiency.

[0069] S102: Input the target sales feature information into a pre-trained sales prediction model, and obtain the average sales in a preset date range and the daily sales corresponding to each date in the preset date range through the sales prediction model.

[0070] The sales forecasting model refers to a model used for sales forecasting, which can be a neural network model. For example, a sales forecasting model is obtained by training with an architecture such as LSTM (Long Short-Term Memory Network). Average sales refers to the average of total sales in a preset date range, which is predicted by the sales forecasting model. Daily sales refers to the sales on each date.

[0071] Exemplarily, the server may input the sales data and sales change data before a preset date interval, as well as the sales data and sales change data in the previous year during the preset date interval into a pre-trained sales prediction model, and process the above-mentioned feature data through the sales prediction model to predict the average sales volume for the preset date interval; and the sales prediction model may also process the above-mentioned feature data to predict the daily sales volume corresponding to each date included in the preset date interval.

[0072] In this way, through the above method, the daily sales corresponding to each date in the preset date range can be effectively predicted, and the predicted average sales volume can be further used to process the daily sales corresponding to each date, and determine the target daily sales corresponding to each date, thereby improving the accuracy of sales forecasts.

[0073] S103: Based on the average sales volume and the daily sales volume, determine the target daily sales volume of the sales point on each date included in the preset date range.

[0074] The target daily sales refers to the daily sales obtained from the final forecast.

[0075] Exemplarily, the server may utilize the predicted average sales volume to perform correction processing on the daily sales volume. For example, if the daily sales volume is greater than the sales volume threshold determined by the average sales volume, it may be considered that there is a prediction error in the daily sales volume, and the daily sales volume may be corrected to obtain the target daily sales volume.

[0076] In this way, processing daily sales through average sales can be used as a fast way to find and iterate the optimal algorithm parameters, thereby improving the accuracy of sales forecasts and improving the efficiency of sales forecasts.

[0077] In some specific examples, when predicting sales during holidays, the server can obtain sales data before and during the holidays (for example, a certain time interval before and during the holidays is taken according to actual conditions) and dimensional data of stores and subcategories from the data center. The sales data obtained here can be sales data for this year and the previous year.

[0078] The target sales feature information is generated mainly through sales fluctuations and trend factors during holidays (for example, the current sales trend of product categories during holidays compared to normal days, such as an increase or decrease).

[0079] For sales forecasts on international holidays, data needs to be highly unified. Otherwise, the special rules and data processing for different regions and holidays will lead to a sudden increase in engineering complexity. By abstracting unified rules to process feature data, we can accurately predict sales on holidays in different regions and improve the accuracy of sales forecasts.

[0080] In this embodiment, by obtaining the target sales characteristic information associated with the preset date interval to be predicted, and predicting through the pre-trained sales prediction model, the average sales of the preset date interval and the daily sales corresponding to each date in the preset date interval can be obtained, and by using the predicted average sales and the predicted daily sales of each date, the target daily sales of each date in the preset date interval can be effectively and accurately determined. In this way, compared with the traditional technology, this embodiment predicts the average sales and daily sales through the sales prediction model, further determines the average sales and daily sales, and determines the target sales predicted for each date, so that the accuracy of daily sales prediction can be provided.

[0081] In an exemplary embodiment, Figure 2A flow chart of the sales forecast model training process provided in the embodiment of the present application is as follows: Figure 2 As shown, in Figure 1 Based on this, the sales forecasting method is further described. The training method of the sales forecasting model includes S201 to S204, wherein:

[0082] S201. Obtain historical sales characteristic information associated with a preset date interval to be predicted and actual historical daily sales of each date included in the preset date interval to be predicted from sales characteristic information of the sales point.

[0083] The historical sales characteristic information may refer to the sales characteristic information of the preset date interval in previous years, for example, the sales characteristic information of the preset date interval in the previous year, and the sales characteristic information of the preset date interval in the previous two years, etc. The actual historical daily sales may refer to the daily sales of each date included in the preset date interval in previous years.

[0084] Exemplarily, the server may determine historical data of a preset date interval to be predicted from the sales feature information of the sales point, that is, as historical sales feature information, and may determine the historical actual daily sales of each date included in the preset date interval.

[0085] For example, the sales data and sales change data of the previous year and the previous two years can be determined as the historical sales feature information of the previous year and the historical sales feature information of the previous two years, respectively, and the actual historical daily sales of each date included in the preset date range of the previous year and the previous two years can be determined. In this way, the sales data of the previous year and the previous two years can be used for model training.

[0086] Optionally, when predicting sales in a preset date interval, the target sales feature information includes sales feature information of the current year before the preset date interval and sales feature information of the preset date interval in the previous year; then, the historical sales feature information used for training should include sales feature information of the preset date interval in the previous year and sales feature information of the preset date interval in the previous two years. That is, data of the previous two years (previous year and previous two years) needs to be obtained for training and verification.

[0087] S202. Input the historical sales feature information into the sales prediction model to be trained, and obtain the predicted historical average sales for a preset date range and the predicted historical daily sales corresponding to each date in the preset date range through the sales prediction model.

[0088] The predicted historical average sales volume refers to the predicted historical average sales volume. The predicted historical daily sales volume refers to the predicted historical daily sales volume for each date.

[0089] Exemplarily, the server processes the input historical sales feature information through the sales forecasting model to be trained, and can predict the historical average sales forecast value for a preset date range and the historical daily sales forecast value corresponding to each date in the preset date range.

[0090] For example, the sales feature information of holidays in the previous year and the sales feature information of holidays in the previous two years are input into the sales forecasting model to predict the historical average sales forecast value of the holiday in the previous year and the historical daily sales forecast value of each date included in the holiday in the previous year.

[0091] Furthermore, the server can correct the historical daily sales forecast value by using the historical average sales forecast value to obtain the historical forecast target daily sales. That is, the historical forecast target daily sales described here refers to the target daily sales obtained by predicting and correcting the historical sales for a preset date range.

[0092] S203: Determine the historical predicted target daily sales volume of the sales point in a preset date range based on the predicted historical average sales volume and the predicted historical daily sales volume.

[0093] The historical forecast target daily sales volume may refer to the daily sales volume forecast value obtained by correcting the daily sales volume forecasted based on the historical sales situation in a preset date range.

[0094] For example, the server can use the average sales volume predicted based on the historical sales situation to correct the daily sales volume predicted based on the historical sales situation to obtain the target daily sales volume predicted value based on the historical sales situation. Furthermore, the sales forecasting model can be optimized and trained using the historical predicted target daily sales volume and the actual historical daily sales volume.

[0095] S204. Train the sales forecasting model according to the historical forecast target daily sales and the actual historical daily sales to obtain a trained sales forecasting model.

[0096] Exemplarily, the server may optimize and train the sales prediction model based on the difference between the historical predicted target daily sales and the actual historical daily sales to obtain optimized model parameters of the sales prediction model, and may obtain a trained sales prediction model based on the optimized model parameters.

[0097] Optionally, the data optimization training of the sales forecasting model may mainly include the optimization of the improvement rate, sales trend control parameters and prediction result specification parameters, and these parameters will be optimized using the SGD algorithm.

[0098] In this embodiment, by utilizing historical sales feature information to predict the sales prediction model to be trained, the sales prediction model can be optimized and trained based on the historical prediction target daily sales and actual historical daily sales obtained by prediction modification, and a sales prediction model that can be used for effective and accurate prediction can be obtained, which is conducive to the implementation of the sales prediction method provided in this application, thereby ensuring that the accuracy of sales prediction is improved.

[0099] In an exemplary embodiment, obtaining target sales characteristic information associated with a preset date interval to be predicted from sales characteristic information of a sales point includes:

[0100] Obtaining current sales characteristic information of the year in which the preset date interval falls, and historical sales characteristic information corresponding to the preset date interval; wherein the historical sales characteristic information is sales characteristic information of the year in which the preset date interval falls;

[0101] According to the current sales characteristic information and the historical sales characteristic information, the target sales characteristic information is obtained.

[0102] The current sales characteristic information refers to the sales characteristic information of the year in which the preset date interval to be predicted is located, that is, the sales characteristic information of the date before the preset date interval. The historical sales characteristic information may refer to the sales characteristic information of the preset date interval in previous years, for example, it may be the sales characteristic information of the preset date interval in the previous year. For example, if the sales characteristic information is holiday A, then the current sales characteristic information may be the sales characteristic information before holiday A, and the historical sales characteristic information may be the sales characteristic information of holiday A in the previous year.

[0103] For example, when the preset date interval is a holiday, the server can obtain sales feature information before the holiday as the current sales feature information. In addition, the server can obtain sales feature information of the holiday in the previous year as the historical sales feature information.

[0104] The server can determine the target sales characteristic information based on the current sales characteristic information and the historical sales characteristic information. That is, the current sales characteristic information and the historical sales characteristic information can be input into the sales prediction model to perform sales prediction.

[0105] In this embodiment, by obtaining the current sales characteristic information of the year in which the preset date interval falls and the historical sales characteristic information corresponding to the preset date interval, the target sales characteristic information can be obtained according to the current sales characteristic information and the historical sales characteristic information. In this way, the data required to be input for the sales forecasting model for forecasting can be effectively determined to ensure the accuracy of the sales forecast.

[0106] In an exemplary embodiment, Figure 3 Schematic diagram of the sales forecasting method provided in the embodiment of the present application Figure 2 ,like Figure 3 As shown, in Figure 1 Based on the above, the steps in the sales forecasting method are described in detail, wherein: obtaining the current sales feature information of the year in which the preset date interval is located, and the historical sales feature information corresponding to the preset date interval, including S301 to S304, wherein:

[0107] S301, obtaining current sales data and current sales change data for the year within the preset date interval and the date before the preset date interval.

[0108] S302: Obtain historical sales data and historical sales change data corresponding to a preset date range.

[0109] S303. Obtain current sales feature information based on current sales data and current sales change data.

[0110] S304. Obtain historical sales feature information based on historical sales data and historical sales change data.

[0111] The current sales data may refer to the sales value before a preset date interval. The current sales change data may refer to the change trend of sales before a preset date interval, such as an increase or decrease, that is, the trend factor may be represented by the change data.

[0112] The historical sales data may refer to the historical sales values ​​within a preset date interval. The historical sales change data may refer to the historical sales change values ​​within a preset date interval. It is understood that the historical sales data and the historical sales change data should correspond to the same time. For example, if the historical sales data is the sales data of the previous year, then the historical sales change data should also be the sales change data of the previous year. As an example, the present application uses the data of the preset date interval of the previous year to perform sales forecasting, and can accurately obtain the forecasted sales.

[0113] Exemplarily, the current sales characteristic information includes current sales data and current sales change data. When the preset date interval is a holiday, the sales characteristic information of the holiday is obtained based on the sales data before the holiday and the sales change data between before and during the holiday.

[0114] The historical sales characteristic information includes historical sales data and historical sales change data. When the preset date interval is a holiday, the sales characteristic information of the holiday of the previous year is obtained based on the sales data of the holiday of the previous year and the sales change data between before and during the holiday of the previous year.

[0115] In this embodiment, the current sales feature information is obtained through the current sales data and the current sales change data, and the historical sales feature information is obtained through the historical sales data and the historical sales change data, which can effectively realize sales forecasting, and ensure that the data features are few, the data construction is fast, and the overfitting risk can be reduced, thereby improving the accuracy of sales forecasting. At the same time, when there are few features and a small number of parameters, the model optimization parameter fitting and result inference speed are very fast, which improves the efficiency of sales forecasting.

[0116] In an exemplary embodiment, Figure 4 Schematic diagram of the sales forecasting method provided in the embodiment of the present application Figure 3 ,like Figure 4 As shown, in Figure 1 Based on the above, the steps in the sales forecasting method are described in detail. In step S101, target sales feature information associated with the preset date interval to be predicted is obtained from the sales feature information of the sales point, including S401 to S402, wherein:

[0117] S401. Obtaining initial sales feature information associated with a preset date range to be predicted from sales feature information of a sales point.

[0118] S402: Perform discrete detection on the initial sales feature information and process the discrete data to obtain the target sales feature information.

[0119] The initial sales feature information may refer to unprocessed feature information directly obtained from sales feature information at the point of sale.

[0120] For example, the server may obtain unprocessed sales feature information associated with the preset date interval to be predicted from the sales feature information of the sales point as the initial sales feature information. The server may perform outlier detection on the obtained initial sales feature information to determine whether there is outlier data, and further perform smoothing or elimination on the outlier data.

[0121] Optionally, the server may process sales trends and sales data with excessive fluctuations, and may use 1.5IQR for outlier detection. After outlier detection, the server may perform data processing such as smoothing or elimination.

[0122] In this embodiment, by performing discrete detection on the initial sales feature information and processing discrete data, the target sales feature information can be obtained, thereby improving the accuracy of the data and further improving the accuracy of the sales forecast.

[0123] In an exemplary embodiment, the method further comprises:

[0124] Get the previous forecast sales of the previous forecast date interval corresponding to the preset date interval, and the actual sales of the previous forecast date interval.

[0125] The sales forecast model is updated using the previous predicted sales and actual sales.

[0126] The previous forecast date interval may refer to the date interval for sales forecasting before the current preset date interval. For example, when the preset date interval is a holiday, the previous forecast date interval may be the time period before the holiday or the previous holiday. The previous forecast sales volume may refer to the sales volume of the previous forecast date interval predicted by the sales forecast model.

[0127] Exemplarily, the server may determine the predicted sales volume for the last forecast date interval predicted by the sales volume forecasting model and the actual sales volume for the last forecast date interval, and optimize and update the sales volume forecasting model.

[0128] For example, the sales forecasting model can be used to predict pre-holiday sales and obtain actual pre-holiday sales. The sales forecasting model can be optimized and updated based on the predicted pre-holiday sales and actual pre-holiday sales. In this way, rolling data updates and forecasts can be achieved, providing better real-time performance for the sales forecasting model, thereby improving the accuracy of model predictions.

[0129] Optionally, the average sales volume in a preset date range and the daily sales volume corresponding to each date in the preset date range are obtained through the sales volume forecasting model, including:

[0130] Through the updated sales forecasting model, the average sales volume in the preset date range and the daily sales volume corresponding to each date in the preset date range are obtained.

[0131] For example, in step S102, the updated sales forecasting model may be used to forecast the average sales in the preset date interval and the daily sales corresponding to each date in the preset date interval, thereby improving the accuracy of sales forecasting for the preset date interval.

[0132] In this embodiment, through the above method, the sales forecast model can be updated in real time, providing better real-time performance for the sales forecast model, thereby improving the accuracy of the model prediction and improving the accuracy of sales forecast for a preset date range.

[0133] In an exemplary embodiment, in an exemplary embodiment, Figure 5 Schematic diagram of the sales forecasting method provided in the embodiment of the present application Figure 4 ,like Figure 5 As shown, in Figure 1Based on the above, the sales forecasting method is further described, wherein the target sales feature information includes sales features of each product category; the method may further include S501 to S503, wherein:

[0134] S501. Obtain the predicted sales volume of each product category in a preset date range through a sales volume prediction model.

[0135] S502: Obtain category sales in a non-preset date range other than the preset date range.

[0136] S503. Determine the predicted sales change of each product category based on the predicted sales volume of the category and the category sales volume.

[0137] Among them, the product category can be the category of goods sold.

[0138] For example, the sales forecast model can be used to predict the category sales corresponding to each product category in a preset date interval. And by determining the category sales in a non-preset date interval, for example, the category sales before the preset date interval, the sales trend of each product category can be further predicted.

[0139] In this embodiment, through steps S501 to S503, the sales trend of product categories at store sales points can be determined. This trend inference can not only predict sales during holidays but also simulate seasonal category changes relatively well, thereby helping to ensure the interaction of product categories.

[0140] In an exemplary embodiment, based on the average sales volume and the daily sales volume, determining the target daily sales volume of the sales point on each date included in the preset date interval includes:

[0141] Get the degree of difference between daily sales and average sales for each date;

[0142] Based on the degree of difference of each date, the target daily sales volume of the sales point on each date included in the preset date range is determined.

[0143] The degree of difference can be represented by a difference value.

[0144] For example, for the daily sales volume on each date, the server may calculate the degree of difference between the daily sales volume and the average sales volume; and based on the degree of difference corresponding to the daily sales volume, the server may determine the target daily sales volume for each date included in the preset date range at the point of sale.

[0145] Optionally, the server may determine a preset difference degree threshold. If the difference degree corresponding to the daily sales volume is greater than the difference degree threshold, the daily sales volume may be smoothed to obtain the corrected daily sales volume for each date, which is used as the target daily sales volume for each date.

[0146] In this embodiment, the daily sales volume of each date can be corrected by the difference between the daily sales volume of each date and the average sales volume to obtain the target daily sales volume of each date, thereby improving the accuracy of the daily sales volume forecast.

[0147] In an exemplary embodiment, Figure 6 A schematic diagram of a sales forecasting method provided in an embodiment of the present application is shown in FIG. Figure 6 As shown, in Figure 1 On this basis, the sales forecasting method is explained in detail. Taking holidays as the preset date interval as an example, the detailed framework of this method includes: feature construction, holiday average sales forecast, holiday daily sales forecast, and algorithm model parameter optimization.

[0148] Feature construction includes: S601, acquisition of store and sub-category feature data; S602, construction of pre-festival features, mid-festival features, and pre-festival and mid-festival interactive features; S603, design and generation of feature rules.

[0149] For example, step S601 is to determine the sales feature information of the sales point. Step S602 is to determine the target sales feature information associated with the preset date range to be predicted. In step S603, it can be a sales forecast for international holidays, which requires a high degree of data uniformity. Otherwise, the engineering complexity will increase sharply due to the special rules and data processing for different regions and holidays. The feature data can be processed by abstracting unified rules.

[0150] The average sales forecast during holidays includes: S604, abnormal data detection and processing; S605, algorithm model parameter optimization training; S606, average sales forecast during holidays.

[0151] Exemplarily, step S604 can be implemented through steps S401 to S402. That is, step S606 can be implemented through step S102 and its corresponding embodiments.

[0152] The daily sales forecast during holidays includes: S607, abnormal data detection and processing; S608, algorithm model parameter optimization training; S609, daily sales forecast during holidays.

[0153] Exemplarily, the steps S607 can be implemented through the steps S401 to S402. S609 can be implemented through the step S102 and its corresponding embodiments.

[0154] Algorithm model parameter optimization includes: S610, gradient descent algorithm (SGD) update parameters.

[0155] Exemplarily, step S610 may be implemented through steps S201 to S204 and their corresponding embodiments.

[0156] It can be understood that the various steps in the framework of this embodiment can be implemented through the above-mentioned various embodiments, and can be referenced to each other, and will not be repeated here.

[0157] In this embodiment, a comprehensive model can be built by combining the sales growth trend during the holidays of the previous year and the sales trend of store categories before the holidays this year. The sales trend of each day during the holidays is constructed, and the sales trend of store categories is calculated. This trend inference can not only predict the sales during the holidays, but also simulate the changes in seasonal categories. A unified entrance for international holidays can be built (to achieve unified data abstraction and processing), and rolling data updates and predictions provide better real-time performance for the algorithm.

[0158] By combining the sales trends of the same holiday period last year to simulate the daily sales during this year's holidays, and building the sales trends of individual products based on this year's sales of individual products, the sales forecast accuracy is improved;

[0159] With fewer data features, data construction is fast and the risk of overfitting can be reduced;

[0160] The embodiments of the present application provide few features and a small number of parameters, so the model optimization parameter fitting and result inference speed are very fast;

[0161] The embodiment of the present application provides a unified entry point for creating international holiday forecasts, abstracts a unified processing and engineering solution at the data level, and provides forecast results based on different underlying data.

[0162] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0163] The sales forecasting device provided in the embodiment of the present application is described below. The sales forecasting device has the same inventive concept as the above-mentioned sales forecasting method. The implementation solution for solving the problem provided by the device is similar to the implementation solution recorded in the above-mentioned method. Therefore, the specific limitations in one or more sales forecasting device embodiments provided below can refer to the limitations on the sales forecasting method above. The sales forecasting device described below and the sales forecasting method described above can be referenced to each other and will not be repeated here.

[0164] In an exemplary embodiment, Figure 7 A schematic diagram of the structure of a sales forecasting device provided in an embodiment of the present application is shown in FIG. Figure 7 As shown, the sales forecasting device 70 includes: a feature acquisition module 710, a forecasting module 720, and a sales determination module 730, wherein:

[0165] The feature acquisition module 710 is used to acquire target sales feature information associated with a preset date interval to be predicted from the sales feature information of the sales point.

[0166] The prediction module 720 is used to input the target sales feature information into a pre-trained sales prediction model, and obtain the average sales in a preset date range and the daily sales corresponding to each date in the preset date range through the sales prediction model.

[0167] The sales volume determination module 730 is used to determine the target daily sales volume of the sales point on each date included in the preset date interval based on the average sales volume and the daily sales volume.

[0168] In an exemplary embodiment, the feature acquisition module is used to obtain the current sales feature information of the year in which a preset date interval occurs, and the historical sales feature information corresponding to the preset date interval; wherein the historical sales feature information is the sales feature information of the year in which a non-preset date interval occurs; and the target sales feature information is obtained based on the current sales feature information and the historical sales feature information.

[0169] In an exemplary embodiment, the feature acquisition module is used to obtain the current sales data and current sales change data of the year in which a preset date interval falls and the dates before the preset date interval; obtain the historical sales data and historical sales change data corresponding to the preset date interval; obtain current sales feature information based on the current sales data and the current sales change data; obtain historical sales feature information based on the historical sales data and the historical sales change data.

[0170] In an exemplary embodiment, the feature acquisition module is used to obtain initial sales feature information associated with a preset date range to be predicted from sales feature information of a sales point; perform discrete detection on the initial sales feature information, and process discrete data to obtain target sales feature information.

[0171] In an exemplary embodiment, the device also includes a sales volume acquisition module and a model update module.

[0172] The sales acquisition module is used to obtain the last predicted sales volume of the last predicted date interval corresponding to the preset date interval, and the actual sales volume of the last predicted date interval. The sales acquisition module is used to update the sales forecast model using the last predicted sales volume and the actual sales volume.

[0173] The forecasting module is used to obtain the average sales volume in a preset date range and the daily sales volume corresponding to each date in the preset date range through the updated sales forecasting model.

[0174] In an exemplary embodiment, the target sales characteristic information includes sales characteristics of each product category. The device also includes a category sales prediction module, a category sales acquisition module and a sales change module.

[0175] The category sales forecasting module is used to obtain the category forecast sales of each product category in a preset date interval through the sales forecasting model. The category sales acquisition module is used to obtain the category sales in a non-preset date interval outside the preset date interval. The sales change module is used to determine the forecast sales change of each product category based on the category forecast sales and category sales.

[0176] In an exemplary embodiment, the sales volume determination module is used to obtain the difference between the daily sales volume and the average sales volume of each date; and determine the target daily sales volume of each date included in the preset date interval of the sales point according to the difference degree of each date.

[0177] In an exemplary embodiment, the device also includes a model training module.

[0178] The model training module is used to obtain historical sales feature information associated with the preset date range to be predicted, and the actual historical daily sales of each date included in the preset date range to be predicted from the sales feature information of the sales point; input the historical sales feature information into the sales prediction model to be trained, and obtain the predicted historical average sales of the preset date range and the predicted historical daily sales corresponding to each date included in the preset date range through the sales prediction model; based on the predicted historical average sales and the predicted historical daily sales, determine the historical predicted target daily sales of the sales point in the preset date range; according to the historical predicted target daily sales and the actual historical daily sales, train the sales prediction model to obtain a trained sales prediction model.

[0179] In an exemplary embodiment, the present application also provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by one or more processors, the one or more processors execute the steps of any sales forecasting method in the above embodiments.

[0180] In an exemplary embodiment, the present application also provides a computer device having a computer program stored therein, wherein when the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of any of the sales forecasting methods in the above embodiments.

[0181] In an exemplary embodiment, the present application further provides a computer program product, including a computer program, which, when executed by a processor, implements any sales forecast as in the above embodiments. Figure 8 As shown, Figure 8 This is a schematic diagram of the internal structure of a computer device provided in an embodiment of the present application. The computer device 800 may be provided as a server. Figure 8 The computer device 800 includes a processing component 802, which further includes one or more processors, and a memory resource represented by a memory 801 for storing instructions executable by the processing component 802, such as an application. The application stored in the memory 801 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 802 is configured to execute the instructions to perform the text recognition method of any of the above embodiments.

[0182] The computer device 800 may further include a power supply component 803 configured to perform power management of the computer device 800, a wired or wireless network interface 804 configured to connect the computer device 800 to a network, and an input / output (I / O) interface 805. The computer device 800 may operate based on an operating system stored in the memory 801, such as Windows Server TM, Mac OS X TM, Unix TM, Linux TM, Free BSD TM, or the like.

[0183] Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0184] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0185] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0186] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can refer to each other.

[0187] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A sales forecasting method, characterized in that: The method comprises: Obtaining target sales characteristic information associated with a preset date interval to be predicted from the sales characteristic information of the sales point; Inputting the target sales feature information into a pre-trained sales forecasting model, and obtaining the average sales in the preset date interval and the daily sales corresponding to each date in the preset date interval through the sales forecasting model; Based on the average sales volume and the daily sales volume, a target daily sales volume of the sales point on each date included in the preset date interval is determined.

2. The method according to claim 1, characterized in that: The step of acquiring target sales characteristic information associated with a preset date interval to be predicted from the sales characteristic information of the sales point includes: Acquire current sales characteristic information of the year in which the preset date interval falls, and historical sales characteristic information corresponding to the preset date interval; wherein the historical sales characteristic information is sales characteristic information of the year not in which the preset date interval falls; The target sales characteristic information is obtained according to the current sales characteristic information and the historical sales characteristic information.

3. The method according to claim 2, characterized in that The obtaining of the current sales characteristic information of the year in which the preset date interval falls and the historical sales characteristic information corresponding to the preset date interval includes: Obtaining current sales data and current sales change data for the year in which the preset date interval falls and the date before the preset date interval; Obtain historical sales data and historical sales change data corresponding to the preset date range; Obtaining the current sales characteristic information according to the current sales data and the current sales change data; The historical sales characteristic information is obtained according to the historical sales data and the historical sales change data.

4. The method according to claim 1, characterized in that: The step of acquiring target sales characteristic information associated with a preset date interval to be predicted from the sales characteristic information of the sales point includes: Acquiring initial sales characteristic information associated with the preset date interval to be predicted from the sales characteristic information of the sales point; The initial sales volume characteristic information is discretely detected and discrete data is processed to obtain the target sales volume characteristic information.

5. The method according to claim 1, characterized in that The method further comprises: Obtaining the last predicted sales volume of the last predicted date interval corresponding to the preset date interval, and the actual sales volume of the last predicted date interval; Using the last predicted sales volume and the actual sales volume, updating the sales volume prediction model; The step of obtaining the average sales volume in the preset date interval and the daily sales volume corresponding to each date in the preset date interval through the sales volume forecasting model includes: The average sales volume in the preset date interval and the daily sales volume corresponding to each date in the preset date interval are obtained through the updated sales volume forecasting model.

6. The method according to claim 1, characterized in that The target sales characteristic information includes sales characteristics of each product category; the method further includes: Obtaining the predicted sales volume of each product category in the preset date range through the sales volume forecasting model; Obtaining sales volume of categories in non-preset date intervals other than the preset date interval; Based on the predicted sales volume of the category and the sales volume of the category, the predicted sales volume change of each of the product categories is determined.

7. The method according to claim 1, characterized in that The determining, based on the average sales volume and the daily sales volume, the target daily sales volume of the sales point on each date included in the preset date interval comprises: Obtain the degree of difference between the daily sales volume on each date and the average sales volume; According to the difference degree of each date, the target daily sales volume of the sales point on each date included in the preset date range is determined.

8. The method according to any one of claims 1 to 7, characterized in that: The training method of the sales forecasting model includes: Acquire, from the sales characteristic information of the sales point, historical sales characteristic information associated with the preset date interval to be predicted, and actual historical daily sales of each date included in the preset date interval to be predicted; Inputting the historical sales feature information into the sales forecasting model to be trained, and obtaining the predicted historical average sales for the preset date interval and the predicted historical daily sales corresponding to each date in the preset date interval through the sales forecasting model; Determine the predicted target daily sales volume of the sales point in the preset date interval based on the predicted historical average sales volume and the predicted historical daily sales volume; The sales forecasting model is trained according to the historical forecast target daily sales and the actual historical daily sales to obtain a trained sales forecasting model.

9. A sales forecasting device, characterized in that: The device comprises: A feature acquisition module, used to acquire target sales feature information associated with a preset date interval to be predicted from sales feature information of a sales point; A prediction module, used to input the target sales feature information into a pre-trained sales prediction model, and obtain the average sales of the preset date interval and the daily sales corresponding to each date in the preset date interval through the sales prediction model; The sales volume determination module is used to determine the target daily sales volume of the sales point on each date included in the preset date interval based on the average sales volume and the daily sales volume.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

Images