Product sales prediction method and device, equipment, storage medium and program product

Abstract

The invention provides a product sales prediction method and device, equipment, a storage medium and a program product, and relates to the field of financial science and technology. The method comprises the steps of obtaining product description information of a to-be-predicted product; according to the product description information, performing matching in a historical product library to determine a plurality of similar products matched with the to-be-predicted product; acquiring historical sales record data of similar products; according to the historical sales record data, re-classifying the similar products to determine product types corresponding to the similar products; determining a target prediction model corresponding to the similar product according to the product type; according to the target prediction model, carrying out sales volume prediction on the similar products to obtain sales volume prediction values of the similar products; and determining the sales prediction value of the to-be-predicted product according to the sales prediction values of all the similar products matched with the to-be-predicted product. According to the method, the accuracy of product sales volume prediction is enhanced, and the prediction efficiency of the product sales volume is improved.

Description

Technical Field

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

[0002] With the continuous development of society, people's living standards are constantly improving. In order to have a strong economic guarantee for their own lives, more and more people begin to choose to obtain financial products. As a result, various financial products have emerged. However, in the fiercely competitive financial market, the sales volume of each financial product varies greatly, making it difficult to accurately and reasonably predict the sales volume of each product. Therefore, product sales forecasting methods have become a promising direction.

[0003] In the prior art, product sales forecasting mainly relies on manual forecasting or single-sequence time series forecasting based on single data.

[0004] However, faced with problems such as large data volume, data sparsity, multiple data forms, diversified user behavior and diversified products, existing technical methods are unable to effectively predict various products in a targeted manner. At the same time, manual prediction methods greatly increase the time and material costs of prediction, and there are technical problems such as low accuracy of product efficiency prediction and poor prediction efficiency. Summary of the invention

[0005] The present application provides a product sales forecasting method, device, equipment, storage medium and program product to solve the technical problems of low accuracy and poor forecasting efficiency of product efficiency forecasting.

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

[0007] Obtain product description information of the product to be predicted;

[0008] According to the product description information, matching is performed in a historical product library to determine a plurality of similar products that match the product to be predicted, wherein the historical product library includes a plurality of historical products;

[0009] Obtain historical sales record data for similar products;

[0010] Reclassify similar products based on historical sales record data to determine the corresponding product types of similar products;

[0011] According to the product type, determine the target prediction model corresponding to similar products;

[0012] According to the target prediction model, sales forecast is conducted for similar products to obtain sales forecast values ​​for similar products;

[0013] The sales forecast value of the product to be predicted is determined based on the sales forecast values ​​of all similar products that match the product to be predicted.

[0014] In a second aspect, the present application provides a product sales forecasting device, comprising:

[0015] The first acquisition module is used to acquire product description information of the product to be predicted;

[0016] A matching module, used to match in a historical product library according to the product description information to determine a plurality of similar products matching the product to be predicted, wherein the historical product library includes a plurality of historical products;

[0017] The second acquisition module is used to obtain historical sales record data of similar products;

[0018] The classification module is used to reclassify similar products based on historical sales record data to determine the corresponding product types of similar products;

[0019] A first determination module is used to determine a target prediction model corresponding to similar products according to the product type;

[0020] A prediction module is used to predict the sales volume of similar products according to the target prediction model to obtain the sales volume prediction value of similar products;

[0021] The second determination module is used to determine the sales forecast value of the product to be predicted based on the sales forecast values ​​of all similar products matching the product to be predicted.

[0022] In a third aspect, an embodiment of the present application provides a product sales forecasting device, including: a memory, a processor;

[0023] The memory stores computer-executable instructions;

[0024] The processor executes the computer-executable instructions stored in the memory, so that the processor performs the method of the first aspect as described above.

[0025] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement the method of the first aspect as described above.

[0026] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, which implements the method of the first aspect above when executed by a processor.

[0027] The present application provides a product sales forecasting method, device, equipment, storage medium and program product, which obtains product description information of the product to be predicted; matches in a historical product library according to the product description information to determine multiple similar products that match the product to be predicted, wherein the historical product library includes multiple historical products; obtains historical sales record data of similar products; reclassifies similar products according to the historical sales record data to determine the product type corresponding to the similar products; determines the target forecasting model corresponding to the similar products according to the product type; forecasts sales for similar products according to the target forecasting model to obtain sales forecast values ​​for similar products; determines the sales forecast value of the product to be predicted according to the sales forecast values ​​of all similar products that match the product to be predicted, thereby on the one hand, by matching multiple products similar to the product to be predicted in the historical product library, and obtaining these similar products; The historical sales record data of the product can use the existing sales data to predict the sales volume of the product to be predicted, thereby improving the reliability of the prediction. On the other hand, reclassifying similar products according to the historical sales record data and determining the product types corresponding to similar products can help reduce the interference between product differences and further improve the accuracy of the prediction. In addition, by determining the target prediction model corresponding to similar products according to the product type, different prediction models can be used for different types of products without human intervention, thus saving a lot of time and material costs, and improving the pertinence and comprehensiveness of product sales prediction. At the same time, the sales forecast value of the product to be predicted is determined based on the sales forecast values ​​of all similar products that match the product to be predicted, which can further smooth the prediction value and reduce the error that may be caused by a single prediction result, thereby achieving the technical effect of enhancing the accuracy of product sales prediction and improving the efficiency of product sales prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0029] Figure 1 A schematic diagram of an application data processing system architecture provided in an embodiment of the present application;

[0030] Figure 2 A flow chart of a product sales forecasting method provided in an embodiment of the present application Figure 1 ;

[0031] Figure 3 A schematic diagram of the principle of the Wide & Deep-GRU prediction model provided in the embodiment of the present application;

[0032] Figure 4 A flow chart of a product sales forecasting method provided in an embodiment of the present application Figure 2 ;

[0033] Figure 5 A schematic diagram of the structure of a neural network model for target embedding provided in an embodiment of the present application;

[0034] Figure 6 A flow chart of a product sales forecasting method provided in an embodiment of the present application Figure 3 ;

[0035] Figure 7 A flow chart of a product sales forecasting method provided in an embodiment of the present application Figure 4 ;

[0036] Figure 8 A flow chart of a product sales forecasting method provided in an embodiment of the present application Figure 5 ;

[0037] Fig. 9 A schematic diagram of the prediction principle of a product sales prediction method provided in an embodiment of the present application;

[0038] Fig.10 A schematic diagram of the structure of a product sales forecasting device provided in an embodiment of the present application;

[0039] Fig.11 A schematic diagram of the structure of a product sales forecasting device provided in an embodiment of the present application.

[0040] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0041] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0042] It should be noted that the product sales forecasting method, device, equipment, storage medium and program product provided in this application can be used in the field of financial technology, and can also be used in any field other than financial technology. The application field of the product sales forecasting method, device, equipment, storage medium and product in this application is not limited.

[0043] As existing technical methods face problems such as large data volume, data sparsity, multiple data forms, diversified user behaviors and diversified products, they are unable to make effective predictions for various products in a targeted manner. At the same time, manual prediction methods greatly increase the time and material costs of prediction, and there are technical problems such as low accuracy of product efficiency prediction and poor prediction efficiency.

[0044] In response to the above problems, the present application provides a product sales forecasting method, device, equipment, storage medium and program product. On the one hand, by matching multiple products similar to the product to be predicted in the historical product library and obtaining the historical sales record data of these similar products, the existing sales data can be used to predict the sales of the product to be predicted, thereby improving the reliability of the prediction. On the other hand, similar products are reclassified according to the historical sales record data to determine the product types corresponding to the similar products, which helps to reduce the interference of differences between products and further improves the accuracy of the prediction. In addition, the target prediction model corresponding to similar products is determined according to the product type, and different prediction models can be used for different types of products without human intervention, thereby saving a lot of time and material costs, and improving the pertinence and comprehensiveness of product sales forecasting. At the same time, the sales forecast value of the product to be predicted is determined based on the sales forecast values ​​of all similar products matched with the product to be predicted, which can further smooth the forecast value and reduce the error that may be caused by a single forecast result, thereby achieving the technical effect of enhancing the accuracy of product sales forecasting and improving the efficiency of product sales forecasting.

[0045] Figure 1 The present invention provides an application data processing system architecture diagram, wherein the application data processing system is a computer device. Figure 1 As shown, the above architecture includes at least one of a data acquisition device 101 , a processing device 102 and a display device 103 .

[0046] It is understandable that the structure illustrated in the embodiment of the present application does not constitute a specific limitation on the application data processing system architecture. In other feasible implementations of the present application, the above architecture may include more or fewer components than shown in the figure, or combine certain components, or split certain components, or arrange the components differently, which can be determined according to the actual application scenario and is not limited here. Figure 1 The components shown may be implemented in hardware, software, or a combination of software and hardware.

[0047] In a specific implementation process, the data acquisition device 101 may include an input / output interface and may also include a communication interface. The data acquisition device 101 may be connected to the processing device via the input / output interface or the communication interface.

[0048] The processing device 102 can obtain product description information of the product to be predicted; match in the historical product library according to the product description information to determine multiple similar products that match the product to be predicted, wherein the historical product library includes multiple historical products; obtain historical sales record data of similar products; reclassify similar products according to the historical sales record data to determine the product type corresponding to the similar products; determine the target prediction model corresponding to the similar products according to the product type; predict the sales volume of similar products according to the target prediction model to obtain the sales volume prediction value of the similar products; determine the sales volume prediction value means of the product to be predicted according to the sales volume prediction values ​​of all similar products that match the product to be predicted.

[0049] The display device 103 may also be a touch display screen or a screen of a terminal device, which is used to receive user instructions while displaying the above content to achieve interaction with the user.

[0050] It should be understood that the above-mentioned processing device can be implemented by a processor reading instructions in a memory and executing the instructions, or it can be implemented by a chip circuit.

[0051] In addition, the network architecture and business 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 in the embodiments of the present application. Ordinary technicians in this field can know that with the evolution of network architecture and the emergence of new business scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0052] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0053] Figure 2 A flow chart of a product sales forecasting method provided in an embodiment of the present application Figure 1 ,like Figure 2 As shown, the product sales forecasting method provided in the embodiment of the present application includes:

[0054] S201, obtaining product description information of the product to be predicted;

[0055] S202: matching in a historical product library according to the product description information to determine a plurality of similar products that match the product to be predicted;

[0056] In this embodiment, the historical product library includes multiple historical products.

[0057] According to the product description information of the product to be predicted, each historical product in the historical product library is matched one by one, and a plurality of similar products matching the product description information of the predicted product are screened out.

[0058] S203, obtaining historical sales record data of similar products;

[0059] In this embodiment, historical sales record data includes basic information, marketing information and sales information, wherein the basic information includes information such as product number, description and category level, the marketing information includes information such as marketing investment, price list, promotion arrangement, etc., and the sales information includes information such as total daily sales, sales through various channels and page views.

[0060] Obtain the historical sales records of each similar product including basic information, marketing information and sales information.

[0061] S204. Reclassify similar products according to historical sales record data to determine the product types corresponding to the similar products;

[0062] Based on the historical sales record data of similar products, continue to reclassify similar products and determine the product type corresponding to each similar product after reclassification.

[0063] S205. Determine a target prediction model corresponding to similar products according to the product type;

[0064] In this embodiment, the target prediction model refers to the Wide&Deep-GRU prediction model.

[0065] According to the product type of each similar product, determine the Wide&Deep-GRU prediction model that matches the similar product.

[0066] Specifically, Figure 3 A schematic diagram of the principle of the Wide & Deep-GRU prediction model provided in the embodiment of the present application, wherein: Figure 3 The GRU_cell in the figure refers to the basic component unit of GRU. Figure 3As shown in the figure, since the data contains time series characteristics, but the processing of continuous information in the Deep network does not take its time series characteristics into consideration, a gated recurrent neural network model (GRU) is added to the Deep network. GRU is an improved version of a recurrent neural network (RNN) and is mainly used to process sequence data. GRU introduces a gating mechanism to solve the gradient vanishing and gradient exploding problems that may occur when the standard RNN processes long sequences, so as to ensure that it can maintain good performance in longer sequences. Therefore, based on the Deep network and GRU, combined with logistic regression (Wide), a Wide&Deep-GRU model is built. This model can express the time series of the input data by introducing GRU, and the activation function of the output layer is changed to ReLU to adapt to the regression prediction problem.

[0067] Parameter settings The hidden layer parameters of the three-layer GRU model are 128, 64, and 64 respectively, and the regularization (Dropout) parameter is set to 0.2. The number of neurons in the three fully connected layers is 32, 32, and 16, and the Dropout parameter is set to 0.15. Xavier initialization and Adam optimizer are used to update the network parameters.

[0068] It should be noted that Figure 3 It is only a reference for displaying the effect and is not a point of improvement and does not affect the protection scope of the embodiments of the present application.

[0069] S206. Predicting sales of similar products according to the target prediction model to obtain sales prediction values ​​of similar products;

[0070] According to the determined target prediction model, sales forecast is performed for the matching similar products to determine the sales forecast value corresponding to the similar products.

[0071] S207: Determine the sales forecast value of the product to be predicted based on the sales forecast values ​​of all similar products that match the product to be predicted.

[0072] According to the sales forecast value of each similar product, the sales forecast value of the product to be predicted that matches it is determined.

[0073] The present embodiment provides a product sales forecasting method, which obtains product description information of a product to be forecasted; matches in a historical product library according to the product description information to determine a plurality of similar products that match the product to be forecasted, wherein the historical product library includes a plurality of historical products; obtains historical sales record data of similar products; reclassifies similar products according to the historical sales record data to determine the product type corresponding to the similar products; determines a target forecasting model corresponding to the similar products according to the product type; forecasts sales for similar products according to the target forecasting model to obtain sales forecast values ​​for the similar products; determines the sales forecast value of the product to be forecasted according to the sales forecast values ​​of all similar products that match the product to be forecasted, thereby, on the one hand, matching a plurality of products similar to the product to be forecasted in the historical product library and obtaining the historical sales record data of these similar products; Record data can be used to predict the sales volume of the product to be predicted, thereby improving the reliability of the prediction. On the other hand, reclassifying similar products according to historical sales record data and determining the product types corresponding to similar products can help reduce the interference between product differences and further improve the accuracy of the prediction. In addition, by determining the target prediction model corresponding to similar products according to product type, different prediction models can be used for different types of products without human intervention, thus saving a lot of time and material costs, and improving the pertinence and comprehensiveness of product sales prediction. At the same time, the sales volume prediction value of the product to be predicted is determined based on the sales volume prediction values ​​of all similar products that match the product to be predicted, which can further smooth the prediction value and reduce the error that may be caused by a single prediction result, thereby achieving the technical effect of enhancing the accuracy of product sales prediction and improving the prediction efficiency of product sales.

[0074] Optional, Figure 4 A flow chart of a product sales forecasting method provided in an embodiment of the present application Figure 2 ,like Figure 4 As shown, based on the above embodiment, this embodiment describes in detail the process of determining the product type corresponding to the similar products, including:

[0075] S401, inputting historical sales record data of similar products into the target embedding neural network model to determine product embedding data of similar products through the target embedding neural network model;

[0076] In this embodiment, the target embedding neural network model is obtained by training the historical sales record data and embedding labels of multiple historical products. The target embedding neural network model refers to an embedded Embedding model, which is used to convert high-dimensional data into a low-dimensional embedding space to retain the original features and semantic information.

[0077] The historical sales record data of similar products are input into a pre-trained target embedding neural network model to obtain product embedding data of similar products output by the target embedding neural network model.

[0078] Specifically, Figure 5 A schematic diagram of the structure of a neural network model for target embedding provided in an embodiment of the present application, such as Figure 5 As shown in the figure, the model includes feature fusion (concat) and two fully connected layers (FC layers), and the two FC layers are passed through regularization (dropout). Taking product Embedding as an example, the input end is composed of product Embedding and product features. After product Embedding and product features are input into the neural network model, feature fusion (concat) is first performed, and then information interaction and regularization (dropout) are performed through two fully connected layers (FC layers), and finally the result is output through the Sigmoid function, and the result range is between 0 and 1. The model output means that if the product Embedding and product features match, the output result is 1, and if they do not match, the output result is 0.

[0079] It should be noted that Figure 5 It is only a reference for displaying the effect and is not a point of improvement and does not affect the protection scope of the embodiments of the present application.

[0080] S402, obtaining historical sales record data of multiple historical products;

[0081] Continue to obtain historical sales record data corresponding to multiple historical products in the above historical product library.

[0082] S403, performing embedded vector coding processing on the historical sales record data of the historical product to obtain product embedded data of the historical product;

[0083] The historical sales record data of each historical product is processed by embedded vectorization encoding to obtain the product embedded data corresponding to each historical product after processing.

[0084] S404, using a density-based clustering algorithm to perform clustering processing on the product embedding data of the historical products to obtain cluster types of the historical products;

[0085] Specifically, the DBCAN algorithm is used to perform clustering processing on the product embedding data of historical products to obtain the clustering type of each historical product after processing.

[0086] S405, training the initial clustering model according to the product embedding data and clustering types of the historical products to obtain a target clustering model;

[0087] The product embedding data and clustering types of historical products are used to train the initial clustering model, and the target clustering model after training is obtained.

[0088] S406: Input the product embedding data of similar products into the target clustering model to obtain product types corresponding to the similar products.

[0089] The product embedding data of similar products are input into the target clustering model to obtain the product types corresponding to the similar products output by the target clustering model.

[0090] A product sales prediction method provided in this embodiment converts high-dimensional and complex historical sales record data of similar products into low-dimensional embedded space data through a target embedded neural network model, which not only reduces the dimension of the data, but also retains the characteristics and semantic information of the original data, providing a more efficient and compact data representation for subsequent processing, and adopts a density-based clustering algorithm to cluster the product embedding data of historical products, which can automatically divide products into different types without manually presetting classification standards or performing complex feature engineering. At the same time, the target clustering model obtained through training can accurately map the product embedding data of new similar products to predefined product types, which not only simplifies the process of determining product types, but also improves the accuracy of prediction. In addition, the target clustering model and the target embedded neural network model obtained through training can be continuously updated and optimized as new data are added, thereby maintaining the accuracy of product type classification and adaptability to market changes, achieving the technical effect of enhancing the accuracy of product sales prediction and improving the prediction efficiency of product sales.

[0091] Optional, Figure 6 A flow chart of a product sales forecasting method provided in an embodiment of the present application Figure 3 ,like Figure 6 As shown, this embodiment, based on the above embodiment, describes in detail the process of determining the target prediction model, wherein the target prediction model includes a first target prediction model and a second target prediction model, and the method includes:

[0092] S601, obtaining an initial recurrent neural network model combining logistic regression and deep neural network;

[0093] S602, determining a training data set of a prediction model corresponding to each cluster type, and training an initial recurrent neural network model combining logistic regression and a deep neural network according to the training data set of the prediction model corresponding to each cluster type, so as to obtain a first target prediction model corresponding to each cluster type;

[0094] In this embodiment, the training data set includes product embedding data and prediction labels of multiple historical products.

[0095] The training data set of the prediction model corresponding to each clustering type is obtained, and the training data set of the prediction model corresponding to each clustering type is input into the initial recurrent neural network model combining logistic regression and deep neural network for training, and the first target prediction model corresponding to each clustering type is obtained after the training is completed.

[0096] S603, determining a first target prediction model corresponding to similar products according to the product type;

[0097] Based on the product type, determine the first target prediction model that matches similar products.

[0098] S604, obtaining an initial lightweight gradient boosting machine learning model;

[0099] In this embodiment, the lightweight gradient boosting machine learning model refers to the LightGBM model.

[0100] Get the initial LightGBM model.

[0101] S605, obtaining product embedding data of multiple historical products, determining prediction labels of the multiple historical products according to a label distribution smoothing algorithm, and using the product embedding data and prediction labels of the historical products as training data sets for a prediction model corresponding to each clustering type;

[0102] In this embodiment, the label distribution smoothing algorithm refers to the LDS (Label Distribution Smoothing) algorithm.

[0103] The product embedding data of multiple historical products are obtained, and the prediction label of each historical product is obtained according to the LDS algorithm. The product embedding data and the prediction label of the historical product are used as the training data set of the prediction model corresponding to each clustering type.

[0104] S606: training the initial lightweight gradient boosting machine learning model according to the training data set of the prediction model corresponding to each cluster type to obtain a second target prediction model corresponding to each cluster type;

[0105] In this embodiment, the training data set includes product embedding data and prediction labels of multiple historical products.

[0106] The training data set of the prediction model corresponding to each cluster type is input into the initial LightGBM model for training, and the second target prediction model corresponding to each cluster type after training is obtained.

[0107] S607: Determine a second target prediction model corresponding to similar products according to the product type.

[0108] Based on the product type, a second target prediction model matching similar products is determined.

[0109] A product sales prediction method provided in this embodiment trains an initial recurrent neural network model that combines logistic regression and deep neural network for each cluster type to obtain a more refined first target prediction model. This refined processing can make more accurate sales predictions for different types of product characteristics, thereby improving the pertinence and comprehensiveness of product efficiency. At the same time, a label distribution smoothing algorithm is used to determine the prediction labels of multiple historical products, which helps to smooth the label distribution and reduce the impact of extreme label values ​​on model training, thereby improving the accuracy and reliability of the prediction. In addition, by introducing a lightweight gradient boosting machine learning model, it is possible to reduce the computing cost while maintaining the prediction performance, thereby achieving the technical effect of enhancing the accuracy of product sales prediction and improving the prediction efficiency of product sales.

[0110] Optional, Figure 7 A flow chart of a product sales forecasting method provided in an embodiment of the present application Figure 4 ,like Figure 7 As shown, based on the above embodiment, this embodiment provides a supplementary description of the process of obtaining sales forecast values ​​of similar products, including:

[0111] S701, predicting sales of similar products according to the first target prediction model to obtain a first sales prediction value of the similar products;

[0112] The first target prediction model is called to perform sales prediction for similar products, and a first sales prediction value output by the first target prediction model is obtained.

[0113] S702: predicting the sales volume of similar products according to the second target prediction model to obtain a second sales volume prediction value of the similar products;

[0114] The second target prediction model is called to perform sales prediction for similar products, and a second sales prediction value output by the second target prediction model is obtained.

[0115] S703: Perform weighted calculation on the first sales forecast value and the second sales forecast value to obtain sales forecasts for similar products.

[0116] The sales forecast value of each similar product is weighted and summed to determine the sales forecast value of the product to be predicted. The weighted summation formula of the sales forecast value is as follows:

[0117]

[0118] in, is the sales forecast value of the product to be predicted, is the predicted value of similar product k, is the weight, which is determined by the matching similarity. The weight formula is as follows:

[0119]

[0120] in, is the weight, x is the product embedding vector, x k is the embedding vector of similar product k, x s is the embedding vector of the top three similar products.

[0121] A product sales forecasting method provided in this embodiment combines the first target model and the second target model to obtain the first sales forecast value and the second sales forecast value corresponding to similar products, thereby reducing the forecast processing time, and using a weighted summation method to determine the sales forecast value of the product to be predicted, thereby further improving the reliability and robustness of the sales forecast, and achieving the technical effect of enhancing the accuracy of product sales forecasting and improving the efficiency of product sales forecasting.

[0122] Optional, Figure 8 A flow chart of a product sales forecasting method provided in an embodiment of the present application Figure 5 ,like Figure 8 As shown, this embodiment provides a supplementary description of the process of determining similar products based on the above embodiment, including:

[0123] S801, extracting keywords from product description information;

[0124] Specifically, the product description information of the product to be predicted is subjected to keyword extraction through the TF-IDF algorithm to obtain corresponding keywords.

[0125] S802: Matching is performed in a historical product library according to the keywords to determine a plurality of similar products that match the product to be predicted.

[0126] According to the above keywords, the keywords of each historical product in the historical product library are matched, and multiple similar products that match the product description information of the predicted product are screened out.

[0127] A product sales forecasting method provided in this embodiment can screen out historical products that are most similar to the product to be forecasted by using a keyword matching method, so as to use the sales data of these similar products to more accurately forecast the sales of the product to be forecasted, which helps to improve the accuracy and reliability of the forecast. At the same time, by extracting keywords and matching them in the historical product library, the calculation complexity is reduced, the timeliness and convenience of the forecast are ensured, and the technical effect of enhancing the accuracy of product sales forecasting and improving the efficiency of product sales forecasting is achieved.

[0128] Fig. 9 A schematic diagram of the prediction principle of a product sales prediction method provided in an embodiment of the present application, such as Fig. 9 As shown in the figure, the Wide&Deep-GRU prediction model can mine effective information between time series and better characterize discrete data. The LightGBM model can mine lag features, differential features, statistical features, and interactive features between data, and then use the weighted fusion method to combine the two models to predict the future sales of products.

[0129] It should be noted that the Wide&Deep-GRU prediction model is related to the category, and different clustering categories may correspond to different Wide&Deep-GRU prediction models.

[0130] In the LightGBM model, we first construct features based on the sliding window, perform lag, difference, sliding statistics and other processing on the time-series feature attributes to characterize the trend and periodicity of the data. Then we build the LightGBM model, use the embedding method to filter features, and use the grid search method to select the optimal model parameters for model training.

[0131] Finally, the Wide&Deep-GRU model is weightedly fused with the LightGBM model. The weight selection of weighted fusion is achieved by linear fitting the prediction results of the training set. The fusion model can take into account the time series characteristics, discrete characteristics, and highly nonlinear characteristic information in the sales data.

[0132] Understandably, Fig. 9 It is only a reference for illustrative display effects and is not intended as an improvement point and does not affect the protection scope of the embodiments of the present application.

[0133] Fig.10 This is a schematic diagram of the structure of a product sales forecasting device provided in an embodiment of the present application. The device of this embodiment can be in the form of software and / or hardware. Figure 8As shown, a product sales forecasting device 1000 provided in an embodiment of the present application includes: a first acquisition module 1001, a matching module 1002, a second acquisition module 1003, a classification module 1004, a first determination module 1005, a prediction module 1006, and a second determination module 1007:

[0134] The first acquisition module 1001 is used to acquire product description information of the product to be predicted;

[0135] A matching module 1002 is used to match the product description information in a historical product library to determine a plurality of similar products that match the product to be predicted, wherein the historical product library includes a plurality of historical products;

[0136] The second acquisition module 1003 is used to acquire historical sales record data of similar products;

[0137] The classification module 1004 is used to reclassify similar products according to the historical sales record data to determine the product types corresponding to the similar products;

[0138] The first determination module 1005 is used to determine the target prediction model corresponding to similar products according to the product type;

[0139] The prediction module 1006 is used to predict the sales volume of similar products according to the target prediction model to obtain the sales volume prediction value of the similar products;

[0140] The second determination module 1007 is used to determine the sales forecast value of the product to be predicted according to the sales forecast values ​​of all similar products matching the product to be predicted.

[0141] In a possible implementation, the classification module 1004 is further configured to:

[0142] Perform embedded vector coding processing on the historical sales record data of similar products to obtain product embedding data of similar products;

[0143] The product embedding data of similar products are input into the target clustering model to obtain the product types corresponding to the similar products, wherein the target clustering model is trained by the product embedding data and clustering types of multiple historical products.

[0144] In a possible implementation, the classification module 1004 is further configured to:

[0145] The historical sales record data of similar products are input into the target embedding neural network model to determine the product embedding data of similar products through the target embedding neural network model, wherein the target embedding neural network model is trained by the historical sales record data and embedding labels of multiple historical products.

[0146] In a possible implementation, the classification module 1004 is further configured to:

[0147] Get historical sales record data for multiple historical products;

[0148] Performing embedded vectorized encoding processing on the historical sales record data of the historical products to obtain product embedded data of the historical products;

[0149] Using density-based clustering algorithm, clustering is performed on the product embedding data of historical products to obtain the clustering types of historical products;

[0150] The initial clustering model is trained according to the product embedding data and clustering types of historical products to obtain the target clustering model.

[0151] In a possible implementation manner, the first determining module 1005 is further configured to:

[0152] Determine a first target prediction model corresponding to similar products according to the product type, wherein the first target prediction model is a recurrent neural network model that combines logistic regression and deep neural network;

[0153] According to the product type, determine the second target prediction model corresponding to the similar product, wherein the second target prediction model is a lightweight gradient boosting machine learning model;

[0154] Accordingly, sales volume forecasting is performed on similar products according to the target forecasting model to obtain sales volume forecasting values ​​of similar products, including:

[0155] According to the first target prediction model, sales forecasting is performed on similar products to obtain a first sales forecast value of the similar products;

[0156] According to the second target prediction model, sales volume prediction is performed on similar products to obtain a second sales volume prediction value of the similar products;

[0157] The first sales forecast value and the second sales forecast value are weightedly calculated to obtain sales forecast values ​​for similar products.

[0158] In a possible implementation manner, the first determining module 1005 is further configured to:

[0159] Get the initial recurrent neural network model that combines logistic regression and deep neural network;

[0160] Determine a training data set of a prediction model corresponding to each clustering type, wherein the training data set includes product embedding data and prediction labels of multiple historical products;

[0161] According to the training data set of the prediction model corresponding to each clustering type, the initial recurrent neural network model combining logistic regression and deep neural network is trained to obtain the first target prediction model corresponding to each clustering type.

[0162] In a possible implementation manner, the first determining module 1005 is further configured to:

[0163] Get an initial lightweight gradient boosting machine learning model;

[0164] Determine a training data set of a prediction model corresponding to each clustering type, wherein the training data set includes product embedding data and prediction labels of multiple historical products;

[0165] According to the training data set of the prediction model corresponding to each clustering type, the initial lightweight gradient boosting machine learning model is trained to obtain the second target prediction model corresponding to each clustering type.

[0166] In a possible implementation manner, the first determining module 1005 is further configured to:

[0167] Get product embedding data for multiple historical products;

[0168] According to the label distribution smoothing algorithm, the predicted labels of multiple historical products are determined.

[0169] In a possible implementation, the matching module 1002 is further configured to:

[0170] Extract keywords from product description information;

[0171] Based on the keywords, a match is performed in the historical product library to determine multiple similar products that match the product to be predicted.

[0172] The product sales forecasting device provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and this embodiment will not be described in detail here.

[0173] Fig.11 This is a schematic diagram of the structure of a product sales forecasting device provided in an embodiment of the present application. Fig.11 As shown, the electronic device 1100 provided in this embodiment includes: at least one processor 1101 and a memory 1102. Optionally, the device 1100 also includes a communication component 1103. The processor 1101, the memory 1102 and the communication component 1103 are connected via a bus.

[0174] In a specific implementation process, at least one processor 1101 executes the computer-executable instructions stored in the memory 1102, so that at least one processor 1101 executes the above method.

[0175] The specific implementation process of the processor 1101 can be found in the above method embodiment, and its implementation principle and technical effect are similar, so this embodiment will not be repeated here.

[0176] In the above embodiments, it should be understood that the processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the invention can be directly implemented as a hardware processor, or can be implemented by a combination of hardware and software modules in the processor.

[0177] The memory may include a high-speed memory (Random Access Memory, RAM), and may also include a non-volatile memory (NVM), such as at least one disk storage.

[0178] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the bus in the drawings of this application is not limited to only one bus or one type of bus.

[0179] The present application also provides a computer program product, including a computer program, which implements the above method when executed by a processor.

[0180] The present application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the above method is implemented.

[0181] The above-mentioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0182] An exemplary readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (Application Specific Integrated Circuits, referred to as: ASIC). Of course, the processor and the readable storage medium can also exist in the device as discrete components.

[0183] The division of units is only a logical function division, and there may be other divisions in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0184] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0185] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0186] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0187] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk and other media that can store program codes.

[0188] Finally, it should be noted that those skilled in the art will readily conceive of other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses or adaptations of the present invention, which follow the general principles of the present invention and include common knowledge or customary technical means in the art not disclosed by the present invention, are not limited to the precise structure described above and shown in the drawings, and may be modified and changed in various ways without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A product sales forecasting method, characterized in that: include: Obtain product description information of the product to be predicted; According to the product description information, matching is performed in a historical product library to determine a plurality of similar products that match the product to be predicted, wherein the historical product library includes a plurality of historical products; Obtaining historical sales record data of similar products; Reclassifying the similar products according to the historical sales record data to determine the product types corresponding to the similar products; Determining a target prediction model corresponding to the similar product according to the product type; According to the target prediction model, sales forecasting is performed on the similar products to obtain sales forecast values ​​of the similar products; The sales volume forecast value of the product to be forecasted is determined according to the sales volume forecast values ​​of all similar products matching the product to be forecasted.

2. The method according to claim 1, characterized in that: The reclassifying the similar products according to the historical sales record data to determine the product types corresponding to the similar products includes: Performing embedded vectorized encoding processing on the historical sales record data of the similar products to obtain product embedded data of the similar products; The product embedding data of the similar products are input into a target clustering model to obtain product types corresponding to the similar products, wherein the target clustering model is obtained by training the product embedding data and clustering types of multiple historical products.

3. The method according to claim 2, characterized in that The embedded vectorized encoding process is performed on the historical sales record data of the similar products to obtain the product embedded data of the similar products, including: The historical sales record data of the similar products are input into the target embedded neural network model to determine the product embedding data of the similar products through the target embedded neural network model, wherein the target embedded neural network model is trained by the historical sales record data and embedded labels of multiple historical products.

4. The method according to claim 2, characterized in that: Before inputting the product embedding data into the target clustering model to obtain the corresponding product types of the similar products, the method further includes: Get historical sales record data of multiple historical products; Performing embedded vectorized encoding processing on the historical sales record data of the historical product to obtain product embedded data of the historical product; Using a density-based clustering algorithm to perform clustering processing on the product embedding data of the historical products to obtain cluster types of the historical products; The initial clustering model is trained according to the product embedding data and clustering types of the historical products to obtain a target clustering model.

5. The method according to claim 4, characterized in that Determining the target prediction model corresponding to the similar product according to the product type includes: Determine, according to the product type, a first target prediction model corresponding to the similar product, wherein the first target prediction model is a recurrent neural network model that combines logistic regression and deep neural network; Determine, according to the product type, a second target prediction model corresponding to the similar product, wherein the second target prediction model is a lightweight gradient boosting machine learning model; Accordingly, the sales volume forecasting of the similar products according to the target forecasting model to obtain the sales volume forecast value of the similar products includes: According to the first target prediction model, sales forecasting is performed on the similar product to obtain a first sales forecast value of the similar product; According to the second target prediction model, sales forecasting is performed on the similar product to obtain a second sales forecast value of the similar product; The first sales forecast value and the second sales forecast value are weightedly calculated to obtain the sales forecast value of the similar product.

6. The method according to claim 5, characterized in that Before determining the first target prediction model corresponding to the similar product according to the product type, the method further includes: Get the initial recurrent neural network model that combines logistic regression and deep neural network; Determine a training data set of a prediction model corresponding to each of the cluster types, wherein the training data set includes product embedding data and prediction labels of a plurality of historical products; According to the training data set of the prediction model corresponding to each of the clustering types, the initial recurrent neural network model combining logistic regression and deep neural network is trained to obtain a first target prediction model corresponding to each of the clustering types.

7. The method according to claim 5, characterized in that Before determining the second target prediction model corresponding to the similar product according to the product type, the method further includes: Get an initial lightweight gradient boosting machine learning model; Determine a training data set of a prediction model corresponding to each of the cluster types, wherein the training data set includes product embedding data and prediction labels of a plurality of historical products; According to the training data set of the prediction model corresponding to each of the clustering types, the initial lightweight gradient boosting machine learning model is trained to obtain a second target prediction model corresponding to each of the clustering types.

8. The method according to claim 6 or 7, characterized in that: The determining of the training data set of the prediction model corresponding to each cluster type includes: Get product embedding data for multiple historical products; According to the label distribution smoothing algorithm, the predicted labels of multiple historical products are determined.

9. The method according to any one of claims 1 to 7, characterized in that: The matching in the historical product library according to the product description information to determine a plurality of similar products matching the product to be predicted includes: Extracting keywords from the product description information; According to the keywords, a match is performed in a historical product library to determine a plurality of similar products that match the product to be predicted.

10. A product sales forecasting device, characterized in that: include: The first acquisition module is used to acquire product description information of the product to be predicted; A matching module, configured to match the product description information in a historical product library to determine a plurality of similar products that match the product to be predicted, wherein the historical product library includes a plurality of historical products; A second acquisition module is used to acquire the historical sales record data of the similar products; A classification module, used for reclassifying the similar products according to the historical sales record data to determine the product types corresponding to the similar products; A first determination module, used to determine a target prediction model corresponding to the similar product according to the product type; A prediction module, used to predict the sales volume of the similar products according to the target prediction model to obtain the sales volume prediction value of the similar products; The second determination module is used to determine the sales forecast value of the product to be predicted according to the sales forecast values ​​of all similar products matching the product to be predicted.

11. A product 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 9.

12. 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 9 when executed by a processor.

13. 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 9 when being executed by a processor.

Images