Product market demand forecasting system and method based on data mining

Abstract

The present application provides a product market demand forecasting system and method based on data mining, which relates to the field of product data mining. It uses artificial intelligence-based data analysis technology to first obtain sales data and order data of different types of products in different historical time periods, then extract features from the acquired data to obtain historical product demand features, and finally generate a product demand forecasting table through a generator using the obtained historical product demand features. In this way, enterprises can better understand the future product demand situation and adjust strategies and plans in a targeted manner to better meet market demand and enhance competitiveness.

Description

Technical Field

[0001] The present application relates to the field of product data mining, and more specifically, to a product market demand forecasting system and method based on data mining. Background Art

[0002] For the decision makers of most product management companies, being able to predict the market demand for various products more timely and making corresponding decisions based on the different market demands of products plays a vital role in the future development of the company. With the continuous optimization and use of data information management systems, many companies' internal databases contain a wealth of historical product sales data. If we can reasonably use Internet technology, based on historical product sales data, and mine data according to the actual market needs of products, we can make corresponding predictions and analyses of various products in the process of future market demand.

[0003] Therefore, a product market demand forecasting solution based on data mining is needed. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present application provides a product market demand forecasting system and method based on data mining.

[0005] A product market demand forecasting system based on data mining, comprising:

[0006] The historical product related data collection module is used to obtain the sales data and order data of different types of products in different historical time periods;

[0007] A historical product related data structuring module, used to arrange the sales data and order data of different types of products in different time periods into a plurality of historical product demand time series input vectors according to the time period dimension;

[0008] A historical product related data feature encoding module, used for performing feature encoding on the plurality of historical product demand time series input vectors to obtain a historical product demand feature vector;

[0009] The product market demand forecast result generating module is used to generate a product demand forecast table based on the historical product demand feature vector.

[0010] A product market demand forecasting method based on data mining, comprising:

[0011] Obtain sales and order data of different types of products in different historical time periods;

[0012] Arranging the sales data and order data of different types of products in different time periods into a plurality of historical product demand time series input vectors according to the time period dimension;

[0013] Performing feature encoding on the plurality of historical product demand time series input vectors to obtain a historical product demand feature vector;

[0014] Based on the historical product demand feature vector, a product demand forecast table is generated.

[0015] The present invention has significant technical effects due to the adoption of the above technical solutions:

[0016] The product market demand forecasting system and method based on data mining provided by the present application adopts data analysis technology based on artificial intelligence, firstly obtains the sales data and order data of different types of products in different historical time periods, then extracts the features of the acquired data to obtain the historical product demand features, and finally generates the product demand forecasting table through the generator by using the obtained historical product demand features. In this way, enterprises can better understand the future product demand situation and adjust strategies and plans in a targeted manner to better meet market demand and enhance competitiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0018] Figure 1 4 is a system block diagram of a product market demand forecasting system based on data mining according to an embodiment of the present application.

[0019] Figure 2 It is a block diagram of a historical product related data feature encoding module in a product market demand forecasting system based on data mining according to an embodiment of the present application.

[0020] Figure 3 It is a block diagram of a historical product demand local feature generation unit in a product market demand prediction system based on data mining according to an embodiment of the present application.

[0021] Figure 4 It is a block diagram of a product demand feature fusion unit in a product market demand prediction system based on data mining according to an embodiment of the present application.

[0022] Figure 5 Flow chart of a method for predicting product market demand based on data mining according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0024] In today's era of information explosion, the importance of market demand forecasting and accurate decision-making has become increasingly prominent for decision makers in most product management companies. With the intensification of market competition and the diversification of consumer demand, being able to quickly and accurately capture market dynamics and make wise decisions based on them has become a key factor in determining the future fate of an enterprise.

[0025] The continuous optimization and wide application of data information management systems have provided enterprises with unprecedented opportunities. Many companies have accumulated a large amount of historical product sales data in their internal databases, which contain rich market information and potential value. However, how to effectively use this data and transform it into valuable market insights is a major challenge facing enterprises. Fortunately, with the rapid development of Internet technology, data mining and analysis have become possible. Enterprises can use advanced algorithms and models to conduct in-depth mining of historical sales data to reveal the market laws and trends hidden behind the data. By grasping these laws, enterprises can more accurately predict the demand for various products in the future market, thereby providing strong support for product strategy adjustments, production planning arrangements and marketing activities. Therefore, a product market demand forecasting solution based on data mining is expected.

[0026] In recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also shown a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields. The development of deep learning and neural networks provides new solutions and solutions for product market demand forecasting.

[0027] Figure 1 FIG. 1 is a system block diagram of a product market demand forecasting system based on data mining according to an embodiment of the present application. Figure 1As shown, in the product market demand forecasting system 100 based on data mining, it includes: a historical product related data collection module 110, which is used to obtain the sales data and order data of different types of products in different historical time periods; a historical product related data structuring module 120, which is used to arrange the sales data and order data of different types of products in different time periods into multiple historical product demand time series input vectors according to the time period dimension; a historical product related data feature encoding module 130, which is used to feature encode the multiple historical product demand time series input vectors to obtain historical product demand feature vectors; a product market demand forecasting result generation module 140, which is used to generate a product demand forecasting table based on the historical product demand feature vectors.

[0028] Specifically, in the technical solution of the present application, the sales data and order data of different types of products in different historical time periods are first obtained. It should be well understood that sales data and order data are a direct reflection of historical market demand. They record the actual sales of various products and customers' ordering intentions in different time periods. These data contain information about market trends, consumer preferences, seasonal changes, and other factors that may affect demand. By analyzing these data, we can gain insight into the historical evolution and laws of market demand. The two types of data obtained are used to train the model when constructing a product demand forecasting model, so that it can learn the complex relationship between market demand and various factors. Through a large amount of data training, the model can identify patterns and trends hidden in the data and predict future market demand accordingly.

[0029] In an embodiment of the present application, a possible implementation method for obtaining sales data and order data of different types of products in different historical time periods may be as follows: sales data of different types of products in different historical time periods may be obtained through an internal sales system. Similarly, order data of different types of products in different historical time periods may be obtained through a supply chain management system.

[0030] Next, the sales data and order data of different types of products in different time periods are arranged into multiple historical product demand time series input vectors according to the time period dimension. It should be well understood that the sales of many products have cyclical characteristics. For example, some products may have a surge in sales during specific seasons or holidays. By arranging the data by time period, the model can more easily identify these cyclical patterns and take them into account when predicting. At the same time, arranging data by time period helps to structure the data so that the model can understand and process it more easily. This structured input data can improve the training efficiency and prediction accuracy of the model. A historical product demand time series input vector represents a sample of product data in a time period. The arrangement in this input vector can be that the first column represents the identification of the first category of products, and different types of products are replaced by different numbers. The second column represents the sales data of the first category of products, the third column represents the order data of the first category of products, and the fourth column represents the identification of the second category of products, and so on.

[0031] Figure 2 FIG. 1 is a block diagram of a historical product related data feature encoding module in a product market demand forecasting system based on data mining according to an embodiment of the present application. Figure 2 As shown, the historical product-related data feature encoding module 130 includes: a historical product demand global feature generation unit 131, used to pass the multiple historical product demand time series input vectors through a product demand feature context encoder to obtain a historical product demand global feature vector; a historical product demand local feature generation unit 132, used to extract local features from a historical product demand time series input matrix formed by arranging the multiple historical product demand time series input vectors to obtain a historical product demand local feature vector; a product demand feature fusion unit 133, used to perform sparsity constraints on the historical product demand global feature vector and the historical product demand local feature vector based on feature implicit expression to obtain the historical product demand feature vector.

[0032] Historical product demands are often not independent, but have certain time series correlations and time series characteristics. In order to capture the correlation and evolution trend between historical product demands, so as to better extract global features and provide more accurate feature information for product market demand forecasting, the multiple historical product demand time series input vectors are passed through the product demand feature context encoder to obtain the historical product demand global feature vector. That is, by context encoding these input vectors, the model can capture these sequence dependencies, so as to more accurately understand the changing trend of market demand. Context encoding not only considers the information of a single time point or a single vector, but also integrates the data of multiple historical time points to form a global feature vector. This helps the model to grasp the dynamic changes of market demand as a whole, rather than being limited to local or short-term data. In this application, the product demand feature context encoder is a context encoder based on a transformer. The Transformer model uses a self-attention mechanism to automatically learn the dependencies between different positions in the input sequence, thereby effectively capturing long-term dependencies and patterns in market demand data. Although (recurrent neural network) RNN can process sequence data, it has the problems of gradient disappearance and low computational efficiency when processing long sequences. The transformer-based encoder overcomes these problems through the self-attention mechanism and is able to process longer sequences and capture more distant dependencies.

[0033] Figure 3 FIG. 1 is a block diagram of a unit for generating local features of historical product demand in a product market demand forecasting system based on data mining according to an embodiment of the present application. Figure 3 As shown, the historical product demand local feature generation unit 132 includes: a matrix local feature extraction subunit 1321, used for arranging the multiple historical product demand time series input vectors into a historical product demand time series input matrix and then passing it through a block-structured product demand local feature encoder to obtain a historical product demand local feature matrix; a feature matrix expansion subunit 1322, used for expanding the historical product demand local feature matrix to obtain the historical product demand local feature vector.

[0034] At the same time, after arranging the multiple historical product demand time series input vectors into a historical product demand time series input matrix, a product demand local feature encoder based on a block structure is used to obtain a historical product demand local feature matrix. Here, it is necessary to first arrange multiple historical product demand time series input vectors to obtain a historical product demand time series input matrix. By arranging multiple vectors into a matrix, it is convenient to process information at multiple time points at the same time, improve the efficiency of data processing and feature extraction, and easily capture the relationship between different time points, so as to better predict future market demand. The original time series input matrix may contain a large number of data points, and direct use for prediction may lead to computational complexity and overfitting problems. Through feature extraction, the original data can be converted into more compact and representative features, so it is necessary to pass the arranged matrix through a product demand local feature encoder. The product demand local feature encoder based on a block structure in this application is a convolutional neural network model based on a block structure. The block structure refers to the way in which the channel attention module and the spatial attention module are stacked. The channel attention module mainly focuses on which channel features are meaningful. By weighting the features of different channels, the model can pay more attention to those features that have an important impact on the prediction. The spatial attention module focuses on the spatial distribution of features. By weighting the features at different positions, the model can pay more attention to those features that are spatially important. The advantage of this combination is that it can take into account both the feature differences between different channels and the differences in the spatial distribution of features. By closely combining the channel attention module and the spatial attention module, the model can more comprehensively capture the key features in the time series data, thereby improving the accuracy of the prediction.

[0035] In an embodiment of the present application, after arranging the multiple historical product demand time series input vectors into a historical product demand time series input matrix, a possible implementation method for obtaining a historical product demand local feature matrix is ​​as follows: using each layer of the block-structured product demand local feature encoder to perform the following on the input data in the forward pass of the layer: performing convolution processing on the input data based on a convolution kernel to obtain a convolution feature map; performing global mean pooling on each feature matrix along the channel dimension of the convolution feature map to obtain a channel feature vector; passing the channel feature vector through a Softmax function to obtain a normalized channel feature vector; using the eigenvalues ​​of each position in the normalized channel feature vector as weights to weight the feature matrix along the channel dimension of the convolution feature map to obtain a channel attention map; performing average pooling and maximum pooling on the channel attention map along the channel dimension to obtain a mean The invention relates to an average feature matrix and a maximum feature matrix; cascading and channel-adjusting the average feature matrix and the maximum feature matrix to obtain a channel feature matrix; using the convolution layer of the spatial attention module to convolutionally encode the channel feature matrix to obtain a convolution feature matrix; passing the convolution feature matrix through the Softmax function to obtain a spatial attention score matrix; performing positional multiplication of the spatial attention score matrix and each feature matrix of the channel attention map along the channel dimension to obtain the channel-spatial attention map; pooling the channel-spatial attention map along the channel dimension to obtain a pooled feature map; performing nonlinear activation on the pooled feature map to obtain an activated feature map; wherein the input of the first layer of the block-structured product demand local feature encoder is the historical product demand time series input matrix, and the output of the last layer of the block-structured product demand local feature encoder is the historical product demand local feature matrix.

[0036] In matrix form, data processing and calculation may involve complex matrix operations, which may increase computational complexity and time cost. By expanding the matrix into a vector, these operations can be simplified, making data processing and calculation more efficient. Therefore, the historical product demand local feature matrix is ​​expanded to obtain the historical product demand local feature vector.

[0037] The global feature vector and the local feature vector each capture different aspects of market demand. The global feature vector is extracted through a context encoder and can fully reflect the overall trend and long-term changes of market demand. The local feature vector is extracted from the original data through models such as convolutional neural networks, focusing on the local details and short-term fluctuations of market demand. In order to comprehensively consider the macro and micro levels of market demand and provide a more comprehensive and accurate market demand forecast, it is necessary to fuse the global feature vector of historical product demand and the local feature vector of historical product demand in the technical solution of this application to obtain the feature vector of historical product demand. There may be complementarity between the global and local feature vectors, and some information that is not obvious in the global features may be highlighted in the local features, and vice versa. By fusing these two vectors, the model can make comprehensive use of these complementary information, thereby improving the accuracy and stability of the prediction. In addition, fusing global and local feature vectors also helps to enhance the robustness of the model. In practical applications, market demand data is often affected by a variety of factors, including seasonality, periodicity, emergencies, etc. Fusion of global and local feature vectors enables the model to better adapt to these complex changes and reduce the errors and uncertainties that may be caused by a single feature vector.

[0038] In particular, in the technical solution of the present application, considering that some features may exist in both the global feature vector of historical product demand and the local feature vector of historical product demand, these overlapping features may provide similar or identical information in the two feature sets. Sometimes, the same information may appear in global and local features in different ways, resulting in the same information being reflected multiple times. In this case, the model may give too high weight to repeated features during learning, thereby affecting the generalization ability and performance of the model. During the fusion process, the feature screening mechanism was not fully considered and implemented. Even if some features appeared repeatedly in global and local features, they were not reasonably filtered, resulting in redundant features entering the final feature vector. Redundant features will increase computational overhead and reduce the efficiency of the model. At the same time, they may also cause overfitting problems, reduce the generalization ability of the model, and affect the accurate prediction of historical product demand. In order to solve this problem, in the technical solution of the present application, the global feature vector of historical product demand and the local feature vector of historical product demand are constrained by sparsity based on implicit expression of features.

[0039] Specifically, Figure 4It is a block diagram of a product demand feature fusion unit in a product market demand forecasting system based on data mining according to an embodiment of the present application. Specifically, the product demand feature fusion unit 133 includes: a first autocorrelation sample covariance matrix generation subunit 1331, which is used to calculate the first autocorrelation sample covariance matrix of the global feature vector of the historical product demand, wherein the eigenvalues ​​of each position in the first autocorrelation sample covariance matrix are the variances between the eigenvalues ​​of the corresponding two positions in the global feature vector of the historical product demand; a second autocorrelation sample covariance matrix generation subunit 1332, which is used to calculate the second autocorrelation sample covariance matrix of the local feature vector of the historical product demand, wherein the eigenvalues ​​of each position in the second autocorrelation sample covariance matrix are the variances between the eigenvalues ​​of the corresponding two positions in the local feature vector of the historical product demand; a covariance matrix principal component analysis processing subunit 1333, which is used to process the first autocorrelation sample covariance matrix and the second autocorrelation sample covariance matrix using the principal component analysis method to obtain the feature implicit representation of the global feature vector of the historical product demand. and the feature implicit expression of the local feature vector of historical product demand; a feature implicit constraint matrix calculation subunit 1334, which is used to use L1 regularization to calculate the feature implicit constraint matrix between the feature implicit expression of the global feature vector of historical product demand and the feature implicit expression of the local feature vector of historical product demand; a vector mapping subunit 1335, which is used to multiply the global feature vector of historical product demand and the local feature vector of historical product demand by the feature implicit constraint matrix respectively to map the global feature vector of historical product demand and the local feature vector of historical product demand to the feature space of the feature implicit constraint matrix to obtain the mapped global feature vector of historical product demand and the mapped local feature vector of historical product demand; a position weighted sum subunit 1336, which is used to calculate the position weighted sum between the mapped global feature vector of historical product demand and the mapped local feature vector of historical product demand to obtain the historical product demand feature vector.

[0040] Here, the sparsity constraints based on implicit expression of features are applied to the global feature vector of historical product demand and the local feature vector of historical product demand, which helps to eliminate irrelevant or redundant features by retaining the most important features, thereby improving feature selectivity and reducing the complexity of the model and the risk of overfitting. In addition, the sparsity constraints based on implicit expression of features can make the fused feature training less sensitive to noise and outliers, thereby improving the robustness of the historical product demand feature vector. At the same time, the historical product demand feature vector after the sparsity constraints based on implicit expression of features contains only a small number of non-zero elements, making it easier to interpret, and also helps to understand the behavior of the model and identify the most important features. And because it only contains a small number of non-zero elements, it requires fewer resources in terms of storage and processing, which can significantly improve computing efficiency.

[0041] More specifically, the covariance matrix principal component analysis processing subunit 1333 is used to: create a Spring Web application; create a data model class, wherein the data model class is used to represent the first autocorrelation sample covariance matrix, the second autocorrelation sample covariance matrix, the feature implicit expression of the historical product demand global feature vector, and the feature implicit expression of the historical product demand local feature vector; write the principal component analysis method in the Spring Web application; create a REST API endpoint, wherein the REST API endpoint is used to allow the client to send the first autocorrelation sample covariance matrix and the second autocorrelation sample covariance matrix; and deploy the Spring Web application on a Web server.

[0042] It should be well understood that there are several reasons for using the Spring Web mechanism to deploy the sparsity constraint feature fusion algorithm based on implicit feature expression: a. Spring Web provides a convenient framework to create RESTful APIs that allow clients to interact with applications through HTTP requests, which makes it easy to deploy algorithms and clients can easily access it from various platforms and devices; b. Spring Web is a lightweight framework that does not have a significant impact on the performance of the application. It is also highly scalable and can handle applications from small to large; c. Spring Web provides built-in security features such as authentication and authorization to protect your application from unauthorized access; d. Spring Web applications can be easily deployed to various web servers such as Apache Tomcat, Nginx, and Cloud Foundry, which makes it simple to deploy algorithms to production environments; e. Spring Web has a large and active community that provides documentation, tutorials, and examples. This makes it easy to learn and use the framework; f. Using the Spring Web mechanism, algorithms can be integrated with other Spring components, such as data access, business logic, and messaging, which can simplify application development and maintenance. In summary, Spring Web Mechanism provides a powerful and easy-to-use platform to deploy sparsity-constrained feature fusion algorithms based on implicit feature expression. It provides advantages such as RESTful API, lightweight and scalability, security, easy deployment, and community support.

[0043] In an embodiment of the present application, a code implementation method for processing the first autocorrelation sample covariance matrix and the second autocorrelation sample covariance matrix using the principal component analysis method to obtain a feature implicit expression of a global feature vector of historical product demand and a feature implicit expression of a local feature vector of historical product demand may be:

[0044]

[0045]

[0046] Finally, the historical product demand feature vector is passed through the product demand generator to obtain a generated result, which is a product demand forecast table. Although the historical product demand feature vector contains rich market demand information, it is not the final forecast result itself. The feature vector is a high-dimensional data representation form, which may be difficult for non-professionals to understand. The product demand forecast table is a more intuitive and easy-to-understand display method that can clearly present the product demand forecast values ​​at different time points. Therefore, by converting the feature vector into a product demand forecast table through the product demand generator, the forecast result can be made easier to understand and use. The product demand generator can generate the corresponding product demand forecast value based on the input feature vector. This generator is usually trained and optimized a lot, and can learn the mapping relationship between the feature vector and the product demand. Therefore, by predicting through the product demand generator, the accuracy and reliability of the prediction results can be ensured.

[0047] In an embodiment of the present application, the historical product demand feature vector is passed through a product demand generator to obtain a generation result, and the generation result is a product demand forecast table. One possible implementation method may be: the structure of the product demand generator here is based on an adversarial generation network, which includes a generation module and an identification module, the generation module is used to generate a product demand forecast table, the identification module is used to calculate the difference between the generated product demand forecast table and the real product demand table, and the network parameters of the generation module are updated through the gradient descent directional propagation algorithm to obtain a generator with the function of generating a product demand forecast table, that is, the product demand generator. The content of the product demand forecast table may include different demand data for different types of goods, and each type of goods is a row of data in the forecast table.

[0048] In summary, the product market demand forecasting system 100 based on data mining according to the embodiment of the present application is explained, which adopts data analysis technology based on artificial intelligence, firstly obtains sales data and order data of different types of products in different historical time periods, then extracts features from the acquired data to obtain historical product demand features, and finally generates a product demand forecasting table through a generator using the obtained historical product demand features. In this way, enterprises can better understand the future product demand situation and adjust strategies and plans in a targeted manner to better meet market demand and enhance competitiveness.

[0049] As described above, the product market demand forecasting system 100 based on data mining according to the embodiment of the present application can be implemented in various terminal devices, such as a server for product market demand forecasting based on data mining. In one example, the product market demand forecasting system 100 based on data mining according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the product market demand forecasting system 100 based on data mining can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the product market demand forecasting system 100 based on data mining can also be one of the many hardware modules of the terminal device.

[0050] Alternatively, in another example, the product market demand forecasting system 100 based on data mining and the terminal device may also be separate devices, and the product market demand forecasting system 100 based on data mining may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0051] Figure 5 FIG. 1 is a flow chart of a method for predicting product market demand based on data mining according to an embodiment of the present application. Figure 5 As shown, in the product market demand forecasting method based on data mining, it includes: S110, acquiring the sales data and order data of different types of products in different historical time periods; S120, arranging the sales data and order data of different types of products in different time periods into multiple historical product demand time series input vectors according to the time period dimension; S130, performing feature encoding on the multiple historical product demand time series input vectors to obtain historical product demand feature vectors; S140, generating a product demand forecasting table based on the historical product demand feature vectors.

[0052] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned product market demand forecasting method based on data mining have been referred to above. Figures 1 to 4 The invention has been introduced in detail in the description of the product market demand forecasting system based on data mining, and therefore, its repeated description will be omitted.

[0053] In summary, the product market demand forecasting method based on data mining according to the embodiment of the present application is explained, which adopts data analysis technology based on artificial intelligence, firstly obtains the sales data and order data of different types of products in different historical time periods, then extracts the features of the acquired data to obtain the historical product demand features, and finally generates the product demand forecast table through the generator by using the obtained historical product demand features. In this way, enterprises can better understand the future product demand situation and adjust strategies and plans in a targeted manner to better meet market demand and enhance competitiveness.

[0054] In the several embodiments provided in this application, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0055] The modules described as separate components may or may not be physically separated, and the components shown as modules 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 modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0056] In addition, each functional module in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0057] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and that the present application can be implemented in other specific forms without departing from the spirit or essential features of the present application.

[0058] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present application is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present application. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

[0059] In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the system claim can also be implemented by one unit or device through software or hardware. The second and other words are used to indicate names, but not to indicate any particular order.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present application and are not intended to limit it. Although the present application has been described in detail with reference to the preferred embodiments, a person of ordinary skill in the art should understand that the technical solution of the present application may be modified or replaced by equivalents without departing from the spirit of the technical solution of the present application.

Claims

1. A product market demand forecasting system based on data mining, characterized in that: include: The historical product related data collection module is used to obtain the sales data and order data of different types of products in different historical time periods; A historical product related data structuring module, used to arrange the sales data and order data of different types of products in different time periods into a plurality of historical product demand time series input vectors according to the time period dimension; A historical product related data feature encoding module, used for performing feature encoding on the plurality of historical product demand time series input vectors to obtain a historical product demand feature vector; A product market demand forecast result generation module, used to generate a product demand forecast table based on the historical product demand feature vector; Wherein, the historical product related data feature coding module includes: A historical product demand global feature generating unit, configured to pass the plurality of historical product demand time series input vectors through a product demand feature context encoder to obtain a historical product demand global feature vector; A historical product demand local feature generating unit, used for extracting local features from a historical product demand time series input matrix formed by arranging the plurality of historical product demand time series input vectors to obtain a historical product demand local feature vector; A product demand feature fusion unit, used for performing sparsity constraints on the historical product demand global feature vector and the historical product demand local feature vector based on feature implicit expression to obtain the historical product demand feature vector; Wherein, the product demand feature fusion unit includes: A first autocorrelation sample covariance matrix generating subunit is used to calculate a first autocorrelation sample covariance matrix of the global feature vector of historical product demand, wherein the eigenvalue of each position in the first autocorrelation sample covariance matrix is ​​the variance between the eigenvalues ​​of two corresponding positions in the global feature vector of historical product demand; A second autocorrelation sample covariance matrix generating subunit is used to calculate a second autocorrelation sample covariance matrix of the local feature vector of historical product demand, wherein the eigenvalue of each position in the second autocorrelation sample covariance matrix is ​​the variance between the eigenvalues ​​of two corresponding positions in the local feature vector of historical product demand; A covariance matrix principal component analysis processing subunit, used to process the first autocorrelation sample covariance matrix and the second autocorrelation sample covariance matrix using a principal component analysis method to obtain a feature implicit expression of a global feature vector of historical product demand and a feature implicit expression of a local feature vector of historical product demand; A feature implicit constraint matrix calculation subunit, used to calculate the feature implicit constraint matrix between the feature implicit expression of the global feature vector of the historical product demand and the feature implicit expression of the local feature vector of the historical product demand using L1 regularization; A vector mapping subunit, used for respectively multiplying the global feature vector of historical product demand and the local feature vector of historical product demand by the feature implicit constraint matrix to map the global feature vector of historical product demand and the local feature vector of historical product demand into the feature space of the feature implicit constraint matrix to obtain the mapped global feature vector of historical product demand and the mapped local feature vector of historical product demand; The position-based weighted sum subunit is used to calculate the position-based weighted sum between the mapped historical product demand global feature vector and the mapped historical product demand local feature vector to obtain the historical product demand feature vector.

2. The product market demand forecasting system based on data mining according to claim 1 is characterized in that: The historical product demand local feature generating unit comprises: A matrix local feature extraction subunit, used for arranging the plurality of historical product demand time series input vectors into a historical product demand time series input matrix and then passing the matrix through a block-structured product demand local feature encoder to obtain a historical product demand local feature matrix; The feature matrix expansion subunit is used to expand the local feature matrix of historical product demand to obtain the local feature vector of historical product demand.

3. The product market demand forecasting system based on data mining according to claim 2 is characterized in that: The product requirement feature context encoder is a converter-based context encoder, and the block-structured product requirement local feature encoder is a block-structured convolutional neural network model.

4. The product market demand forecasting system based on data mining according to claim 3 is characterized in that: The covariance matrix principal component analysis processing subunit is used for: Create a Spring web application; Creating a data model class, wherein the data model class is used to represent the first autocorrelation sample covariance matrix, the second autocorrelation sample covariance matrix, the feature implicit expression of the global feature vector of the historical product demand, and the feature implicit expression of the local feature vector of the historical product demand; Writing the principal component analysis method in the Spring Web application; Creating a REST API endpoint, wherein the REST API endpoint is used to allow a client to send the first autocorrelation sample covariance matrix and the second autocorrelation sample covariance matrix; Deploy the Spring Web application on a Web server.

5. The product market demand forecasting system based on data mining according to claim 4 is characterized in that: The product market demand forecast result generating module is used to: pass the historical product demand feature vector through the product demand generator to obtain a generation result, and the generation result is a product demand forecast table.

6. A product market demand forecasting method based on data mining, characterized in that: include: Obtain sales and order data of different types of products in different historical time periods; Arranging the sales data and order data of different types of products in different time periods into a plurality of historical product demand time series input vectors according to the time period dimension; Performing feature encoding on the plurality of historical product demand time series input vectors to obtain a historical product demand feature vector; Based on the historical product demand feature vector, generate a product demand forecast table; The process of performing feature encoding on the plurality of historical product demand time series input vectors to obtain a historical product demand feature vector includes: Passing the plurality of historical product demand time series input vectors through a product demand feature context encoder to obtain a historical product demand global feature vector; Performing local feature extraction on a historical product demand time series input matrix formed by arranging the plurality of historical product demand time series input vectors to obtain a local feature vector of historical product demand; Performing sparsity constraints based on implicit feature expression on the global feature vector of historical product demand and the local feature vector of historical product demand to obtain the feature vector of historical product demand; The step of performing sparsity constraints based on implicit feature expression on the global feature vector of historical product demand and the local feature vector of historical product demand to obtain the feature vector of historical product demand includes: Calculating a first autocorrelation sample covariance matrix of the global feature vector of historical product demand, wherein the eigenvalue of each position in the first autocorrelation sample covariance matrix is ​​the variance between the eigenvalues ​​of two corresponding positions in the global feature vector of historical product demand; Calculating a second autocorrelation sample covariance matrix of the local feature vector of historical product demand, wherein the eigenvalue of each position in the second autocorrelation sample covariance matrix is ​​the variance between the eigenvalues ​​of two corresponding positions in the local feature vector of historical product demand; Using principal component analysis to process the first autocorrelation sample covariance matrix and the second autocorrelation sample covariance matrix to obtain a feature implicit expression of a global feature vector of historical product demand and a feature implicit expression of a local feature vector of historical product demand; Using L1 regularization to calculate the implicit feature constraint matrix between the implicit feature expression of the global feature vector of the historical product demand and the implicit feature expression of the local feature vector of the historical product demand; Multiplying the global feature vector of historical product demand and the local feature vector of historical product demand by the feature implicit constraint matrix respectively to map the global feature vector of historical product demand and the local feature vector of historical product demand into the feature space of the feature implicit constraint matrix to obtain the mapped global feature vector of historical product demand and the mapped local feature vector of historical product demand; The position-weighted sum between the mapped historical product demand global feature vector and the mapped historical product demand local feature vector is calculated to obtain the historical product demand feature vector.

7. The product market demand forecasting method based on data mining according to claim 6 is characterized in that: Based on the historical product demand feature vector, a product demand forecast table is generated, including: The historical product demand feature vector is passed through a product demand generator to obtain a generation result, and the generation result is a product demand forecast table.

Images