Digital commodity order data analysis system and method based on big data

Abstract

The invention relates to the field of order data analysis, and particularly discloses a digital commodity order data analysis system and method based on big data, and the method comprises the steps: firstly obtaining digital commodity order purchase channel information, digital commodity order data and user behavior logs collected by a database, and then carrying out the deep learning technology, and feature extraction and correlation analysis are carried out on the three items, and finally, a personalized user purchase correlation analysis commodity recommendation list report is generated through a generator, so that the correlation and attraction of the commodities are improved, the user resource upper limit is deeply matched, and the business growth and the user satisfaction are promoted.

Description

Technical Field

[0001] The present application relates to the field of order data analysis, and more specifically, to a digital commodity order data analysis system and method based on big data. Background Art

[0002] With the rapid development of Internet technology, the Internet has penetrated into people's study, work and life, and has become an important tool for handling daily affairs. The instant exchange and sharing of digital goods has broken through geographical restrictions and greatly improved transaction efficiency. With the rise of various applications, users have become the core objects of services.

[0003] However, current technologies mainly focus on recording user order data, but lack in-depth analysis and processing of these data, which makes it impossible to fully understand the actual use of resources and user behavior patterns, and thus the applicability of resources.

[0004] Therefore, a digital commodity order data analysis system and method based on big data is desired. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed. The embodiment of this application provides a digital commodity order data analysis system and method based on big data, which first obtains digital commodity order purchase channel information, digital commodity order data and user behavior logs collected by the database, and then uses deep learning technology to perform feature extraction and association analysis on the three, and finally generates a personalized user purchase association analysis commodity recommendation list report through a generator, thereby improving the relevance and attractiveness of the commodity, deeply matching the user resource upper limit, and then promoting business growth and user satisfaction.

[0006] According to one aspect of the present application, a digital commodity order data analysis system based on big data is provided, which includes:

[0007] A digital commodity order data acquisition module is used to acquire digital commodity order purchase channel information, digital commodity order data and user behavior logs collected from the database;

[0008] A digital commodity order data extraction module, used to extract an optimized digital commodity order multimodal association feature vector and a user behavior log semantic understanding feature vector from the digital commodity order purchase channel information collected by the database, the digital commodity order data and the user behavior log;

[0009] The associated product recommendation report generation module is used to generate a personalized user purchase association analysis product recommendation list report based on the optimized digital product order multimodal association feature vector and the user behavior log semantic understanding feature vector.

[0010] According to another aspect of the present application, a method for analyzing digital commodity order data based on big data is provided, which includes:

[0011] Obtain digital product order purchase channel information, digital product order data and user behavior logs collected from the database;

[0012] Extracting an optimized digital commodity order multimodal association feature vector and a user behavior log semantic understanding feature vector from the digital commodity order purchase channel information collected by the database, the digital commodity order data and the user behavior log;

[0013] Based on the optimized digital commodity order multimodal association feature vector and the user behavior log semantic understanding feature vector, a personalized user purchase association analysis commodity recommendation list report is generated.

[0014] Compared with the prior art, the present application provides a digital commodity order data analysis system and method based on big data, which first obtains digital commodity order purchase channel information, digital commodity order data and user behavior logs collected by the database, and then uses deep learning technology to perform feature extraction and association analysis on the three, and finally generates a personalized user purchase association analysis product recommendation list report through a generator, thereby improving the relevance and attractiveness of the products, deeply matching the user resource upper limit, and then promoting business growth and user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0016] Figure 1 4 is a block diagram of a digital commodity order data analysis system based on big data according to an embodiment of the present application.

[0017] Figure 2 4 is a block diagram of a digital commodity order data extraction module in a digital commodity order data analysis system based on big data according to an embodiment of the present application.

[0018] Figure 3 This is a block diagram of a module for generating a recommendation report for associated products in a digital product order data analysis system based on big data according to an embodiment of the present application.

[0019] Figure 4 The present invention is a flowchart of a method for analyzing digital commodity order data based on big data according to an embodiment of the present application.

[0020] Figure 5 is a block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0021] 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.

[0022] Figure 1 : is a block diagram of a digital commodity order data analysis system based on big data according to an embodiment of the present application. Figure 1 As shown, according to the big data-based digital commodity order data analysis system 100 of the embodiment of the present application, it includes: a digital commodity order data acquisition module 110, which is used to obtain digital commodity order purchase channel information, digital commodity order data and user behavior logs collected by the database; a digital commodity order data extraction module 120, which is used to extract optimized digital commodity order multimodal association feature vectors and user behavior log semantic understanding feature vectors from the digital commodity order purchase channel information collected by the database, the digital commodity order data and the user behavior log; an associated commodity recommendation report generation module 130, which is used to generate a personalized user purchase associated analysis commodity recommendation list report based on the optimized digital commodity order multimodal association feature vector and the user behavior log semantic understanding feature vector.

[0023] In the above-mentioned digital commodity order data analysis system 100 based on big data, the digital commodity order data acquisition module 110 is used to obtain the digital commodity order purchase channel information, digital commodity order data and user behavior log collected by the database. It should be understood that with the rapid development of Internet technology, it has been deeply integrated into people's study, work and daily life, and has become a key tool for handling various affairs. The instant exchange and sharing of digital commodities has broken through geographical restrictions and greatly improved the efficiency of transactions. With the popularization of various applications, users have gradually become the center of services. However, current technologies mainly focus on recording users' order data, while in-depth analysis and processing of these data are relatively insufficient. This limitation makes it difficult to fully understand the actual use of resources and users' behavior patterns, so that it is impossible to accurately grasp the applicability of resources. Therefore, in the technical solution of this application, by obtaining the digital commodity order purchase channel information, digital commodity order data and user behavior log collected by the database, and combining deep learning technology, a personalized user purchase association analysis report is generated to improve the relevance and attractiveness of the goods, accurately match the upper limit of user demand, thereby promoting business growth and improving user satisfaction.

[0024] Specifically, obtaining digital product order purchase channel information, digital product order data, and user behavior logs collected by the database is the basis for building an accurate and personalized recommendation system. Purchase channel information reveals the user's purchase path and preferences, including the device, platform, or sales channel selected by the user, which helps to identify which channels are most effective and which specific paths may affect purchase decisions. Digital product order data provides specific product purchase records, including key information such as product type, quantity, and price. User behavior logs record the detailed behavior of users during browsing and interaction, such as clicks, searches, and browsing history. Through comprehensive analysis of these data, the system can extract valuable features and generate a user's purchase tendency model, thereby providing more accurate personalized recommendations.

[0025] In the above-mentioned digital commodity order data analysis system 100 based on big data, the digital commodity order data extraction module 120 is used to extract the optimized digital commodity order multimodal association feature vector and the user behavior log semantic understanding feature vector from the digital commodity order purchase channel information collected by the database, the digital commodity order data and the user behavior log. In this way, the efficiency of data processing can be improved, so that the recommendation system can capture the changes and trends of user needs in a more accurate way, thereby helping to generate more targeted recommendations, reduce recommendation errors, and improve user experience.

[0026] Figure 2 FIG. 1 is a block diagram of a digital commodity order data extraction module in a digital commodity order data analysis system based on big data according to an embodiment of the present application. Figure 2 As shown, in a specific embodiment of the present application, the digital commodity order data extraction module 120 includes: a purchase channel information feature extraction unit 121, which is used to extract features from the digital commodity order purchase channel information collected by the database to obtain a digital commodity order purchase channel semantic feature vector; a commodity order data feature extraction unit 122, which is used to extract features from the digital commodity order data to obtain a digital commodity order semantic association feature vector; a commodity order multimodal feature aggregation unit 123, which is used to perform feature aggregation on the digital commodity order purchase channel semantic feature vector and the digital commodity order semantic association feature vector to obtain the optimized digital commodity order multimodal association feature vector; a user behavior log feature extraction unit 124, which is used to extract features from the user behavior log to obtain the user behavior log semantic understanding feature vector.

[0027] It should be understood that purchase channel information not only includes the specific channels selected by users (such as websites, mobile applications, physical stores, etc.), but may also involve the devices used, the time period of access, transaction frequency, etc. This information provides a valuable data source for analyzing users' purchasing habits, preferences, and channel effectiveness. The key to feature extraction is to convert these raw data into semantic feature vectors with practical application value. This process includes extracting and encoding relevant features from purchase channel information, such as the type of channels visited by users, the conversion rate of channels, the purchasing power of channels, etc. Feature extraction of purchase channel information can not only simplify complex channel data into a processable form, but also deeply understand the relationship between user behavior and channels. This in-depth analysis provides a solid data foundation for personalized recommendations, enabling the system to better identify and meet user needs, and ultimately improve user satisfaction and the effectiveness of the recommendation system.

[0028] Furthermore, digital product order data usually contains basic order information, such as product ID, purchase time, purchase quantity, price, and user ID. Although these data are structured, in order to fully utilize their value in personalized recommendations and market analysis, they need to be subjected to in-depth feature extraction and semantic association analysis. By analyzing users' historical order data, we can build a user preference model and identify users' interests and needs for certain products, which is crucial for accurate recommendations.

[0029] Furthermore, feature aggregation is performed on the digital product order purchase channel semantic feature vector and the digital product order semantic association feature vector to obtain the optimized digital product order multimodal association feature vector, in order to comprehensively utilize information from different data sources, so as to more comprehensively and accurately capture the complex relationship between users and products. This aggregation process can effectively integrate multiple information, provide deeper insights, and improve the performance and accuracy of the recommendation system. Among them, the digital product order purchase channel semantic feature vector reflects the purchase channel selected by the user and the impact of these channels on the purchase decision. The digital product order semantic association feature vector focuses on the specific product information in the order data, such as product type, purchase time, price, etc., and shows the interaction pattern and purchase trend between the product and the user. Through feature aggregation, these two types of feature vectors are integrated to form a multimodal association feature vector. This aggregation can comprehensively consider the user's purchase channel and specific product information, so as to capture richer user behavior patterns.

[0030] In particular, feature extraction of user behavior logs is to transform various interactive behaviors of users in the system into meaningful feature representations, so as to deeply understand the needs, preferences and behavior patterns of users. User behavior logs usually record the operation history of users on the platform, including page browsing, clicks, searches, purchases, dwell time and other behavior data. Although these log information itself can provide users' operation trajectories, in order to extract useful insights from them, these data must be converted into structured feature vectors for further analysis and application.

[0031] In a specific embodiment of the present application, the purchase channel information feature extraction unit 121 includes: passing the digital commodity order purchase channel information collected from the database through an order purchase channel semantic encoder based on a bidirectional long short-term memory neural network to obtain multiple digital commodity order purchase channel input vectors; passing the multiple digital commodity order purchase channel input vectors through a digital commodity order purchase channel information semantic feature extractor based on a convolutional neural network to obtain the digital commodity order purchase channel semantic feature vector.

[0032] It should be understood that the purchase channel information of digital goods orders contains the data of users purchasing through different channels (such as websites, mobile applications, social media, etc.). These data not only include the specific channels of purchase, but may also involve detailed information such as visit time, frequency, conversion rate, etc. Since this information is time-series and context-dependent, traditional feature extraction methods often find it difficult to fully explore its potential semantic associations. Therefore, it is a more effective strategy to use a bidirectional long short-term memory neural network (Bi-LSTM) for encoding. Bi-LSTM is an improved LSTM network that can better capture long-term dependencies and contextual relationships in the data by simultaneously considering the forward and backward information of time series data. This bidirectional processing method enables the model to capture complex patterns and relationships in channel data, such as the behavioral patterns of users on different channels and the trends of these behaviors over time.

[0033] Further, multiple input vectors containing rich contextual information have been generated using an order purchase channel semantic encoder based on a bidirectional long short-term memory neural network (Bi-LSTM). These vectors can capture the time series features and complex semantic relationships in the channel information, but still require further processing to extract more representative and effective features. Convolutional neural networks (CNNs), especially their ability to process high-dimensional features and capture local patterns, provide an effective solution for this process. By applying convolutional layers, CNNs can slide convolution kernels on input vectors, automatically extract local features, and build higher-level feature representations on this basis. Specifically, each layer of the digital commodity order purchase channel information semantic feature extractor based on the convolutional neural network performs convolution processing, mean pooling processing based on local feature matrices, and nonlinear activation processing on the input data in the forward pass of the layer to output the digital commodity order purchase channel semantic feature vector by the last layer of the digital commodity order purchase channel information semantic feature extractor based on the convolutional neural network, wherein the input of the digital commodity order purchase channel information semantic feature extractor based on the convolutional neural network is the multiple digital commodity order purchase channel input vectors.

[0034] In a specific embodiment of the present application, the commodity order data feature extraction unit 122 includes: passing the digital commodity order data through a converter-based digital commodity order context encoder to obtain multiple digital commodity order feature vectors; and cascading the multiple digital commodity order feature vectors into the digital commodity order semantic association feature vector.

[0035] It should be understood that digital commodity order data usually contains a variety of information such as order time, commodity category, price, user information, etc. The relevance and contextual relationship of this information are crucial to understanding the user's purchasing behavior. However, traditional sequence processing methods often find it difficult to effectively handle long-distance dependencies and complex contextual relationships in the data. At this time, the converter model can play its advantages. The converter allows the model to focus on all positions in the input sequence at the same time when processing data through the self-attention mechanism (Self-Attention), so that the global relationship between different parts of the data can be captured when modeling. Specifically, the Bert model of the converter-based digital commodity order context encoder is used to perform context semantic encoding based on the global time series on each digital commodity order data item in the digital commodity order data to obtain the multiple digital commodity order feature vectors.

[0036] Furthermore, multiple digital product order feature vectors are concatenated into a digital product order semantic association feature vector in order to comprehensively consider the unique features of each order and the relationship between them, thereby generating a more comprehensive and integrated feature representation. Each digital product order feature vector contains specific information about the order, such as product type, price, purchase time, user behavior, etc. These feature vectors extract different aspects of feature information from the order data through different processing steps (such as through convolutional neural networks or transformer-based encoders). However, a single feature vector can often only reflect one aspect of the order data and may not fully display the complex associations between orders and the overall behavior patterns of users. By concatenating multiple feature vectors together, feature information from different sources can be integrated into a unified representation. This cascading operation splices the contents of each feature vector together to form a new, richer feature vector. This semantic association feature vector not only integrates the independent features of each order, but also better captures the relationship between orders and the overall trend of user behavior.

[0037] In a specific embodiment of the present application, the commodity order multimodal feature aggregation unit 123 includes: associating the digital commodity order purchase channel semantic feature vector and the digital commodity order semantic association feature vector to obtain a digital commodity order multimodal association feature matrix; passing the digital commodity order multimodal association feature matrix through a digital commodity order multimodal feature encoder based on a deep convolutional neural network to obtain a digital commodity order multimodal association feature vector; and performing linear signal modulation based on fine-grained features on the digital commodity order multimodal association feature vector based on the semantic understanding feature vector of the user behavior log to obtain an optimized digital commodity order multimodal association feature vector.

[0038] It should be understood that the digital commodity order purchase channel semantic feature vector usually captures the user's purchase behavior and preferences on different channels (such as websites, mobile applications, social media, etc.) from the perspective of the user's purchase channel. On the other hand, the digital commodity order semantic association feature vector starts from the perspective of order content and context, and contains detailed information about the order, such as product type, price, purchase time, and user's purchase habits. By associating these two types of feature vectors, the channel information can be fused with the order content information to form a multimodal association feature matrix. This matrix can not only comprehensively reflect the user's purchase behavior pattern, but also reveal the potential relationship between different purchase channels and order content.

[0039] Furthermore, the multimodal association feature matrix of digital product orders integrates information from different data sources (such as purchase channels and order content), which can fully reflect the user's purchasing behavior and order characteristics. However, these multimodal data usually have complex intrinsic structures and potential patterns, which require further processing to effectively extract useful features. Deep convolutional neural networks (CNNs) have significant advantages in processing high-dimensional data and capturing local features, especially when processing data with spatial or temporal structures. CNN is able to extract local patterns in data through convolutional layers and build higher-level feature representations layer by layer. In this process, the multimodal feature encoder based on deep CNN deeply processes the information in the multimodal association feature matrix by applying multiple convolutional layers. Specifically, each layer of the multimodal feature encoder for digital commodity orders based on a deep convolutional neural network performs the following on the input data in the forward pass of the layer: based on the convolution kernel, convolution processing is performed on the input data to generate a convolutional feature map; global mean pooling processing is performed on the convolutional feature map based on a feature matrix to generate a pooled feature map; and nonlinear activation is performed on the feature values ​​at each position in the pooled feature map to generate an activated feature addition map; wherein the output of the last layer of the multimodal feature encoder for digital commodity orders based on a deep convolutional neural network is the optimized multimodal association feature vector of digital commodity orders, the input from the second layer to the last layer of the multimodal feature encoder for digital commodity orders based on a deep convolutional neural network is the output of the previous layer, and the input of the multimodal feature encoder for digital commodity orders based on a deep convolutional neural network is the multimodal association feature matrix of digital commodity orders.

[0040] In particular, in the technical solution of the present application, the multimodal correlation feature vector of digital commodity orders integrates purchase channel information, order data and the complex relationship between them, involves different types of data processing and feature extraction, has a more complex distribution in high-dimensional space, and emphasizes the relationship between different data dimensions. In contrast, the semantic understanding feature vector of user behavior log is generated through word segmentation and embedding layers, mainly reflecting the semantic information of user behavior. These log data usually have strong local features, and their semantic expression is more intuitive and direct, the feature dimension may be low, and the focus is on the immediate and short-term patterns of user behavior. This difference in feature source and processing method leads to inconsistent fine-grained distribution of multimodal correlation feature vectors of digital commodity orders and semantic understanding feature vectors of user behavior logs in feature space. Specifically, digital commodity order features involve multiple complex patterns and associations, while user behavior log features focus more on immediate behavior patterns. This mismatch will cause problems in the fusion process, because on the boundary of the feature manifold, the mapping of complex order features and simplified user behavior features in high-dimensional space may be inconsistent, resulting in abnormal stacking of feature vectors generated by the fused product recommendation list report, affecting the performance of the recommendation system and the accuracy of the recommendation results. Therefore, in the technical solution of the present application, based on the semantic understanding feature vector of the user behavior log, the digital commodity order multimodal association feature vector is subjected to linear signal modulation based on fine-grained features to obtain an optimized digital commodity order multimodal association feature vector.

[0041] Among them, based on the user behavior log semantic understanding feature vector, the digital commodity order multimodal association feature vector is subjected to linear signal modulation based on fine-grained features to obtain an optimized digital commodity order multimodal association feature vector, including: calculating the position difference between the digital commodity order multimodal association feature vector and the user behavior log semantic understanding feature vector to obtain a digital commodity order shift position difference representation vector; after cascading the digital commodity order multimodal association feature vector and the user behavior log semantic understanding feature vector into a digital commodity order joint feature vector, the digital commodity order joint feature vector is input into a one-dimensional convolutional layer to obtain a digital commodity order shift convolutional coding representation vector; after using a first digital commodity order linear embedding weight matrix to perform digital commodity order linear embedding modulation encoding on the digital commodity order multimodal association feature vector into a digital commodity order multimodal association feature vector linear modulation vector, calculating The digital commodity order multimodal association feature vector linear modulation vector and the user behavior log semantic understanding feature vector are differentially positioned to obtain a digital commodity order shift modulation differential representation vector; the digital commodity order shifted position differential representation vector is used as a value vector, the digital commodity order shifted convolutional coding representation vector and the digital commodity order shift modulation differential representation vector are used as a query vector and a key vector, and the digital commodity order shifted position differential representation vector, the digital commodity order shifted convolutional coding representation vector and the digital commodity order shift modulation differential representation vector are input into a shift corrector based on an imitation transformer structure to obtain a digital commodity order multi-scale hierarchical fusion shift modulation weight vector; the position point multiplication between the digital commodity order multi-scale hierarchical fusion shift modulation weight vector and the digital commodity order multimodal association feature vector is calculated to obtain the optimized digital commodity order multimodal association feature vector.

[0042] The following optimization formula is used to express that the linear signal modulation based on the fine-grained features of the digital commodity order multimodal association feature vector is performed based on the semantic understanding feature vector of the user behavior log to obtain the optimized digital commodity order multimodal association feature vector;

[0043] Wherein, the optimization formula is:

[0044]

[0045] V 2d =Conv 1D ([V1; V2])

[0046]

[0047] V1'=V1⊙V d

[0048] Among them, V1d represents the position difference representation vector of the digital commodity order shift, V1 represents the multimodal association feature vector of the digital commodity order, V2 represents the semantic understanding feature vector of the user behavior log, V 2d Represents the shift convolutional coding representation vector of the digital product order, V 3d represents the digital commodity order shift modulation differential representation vector, W1 represents the first digital commodity order linear embedding weight matrix, b1 represents the digital commodity order modulation bias vector, V d represents the multi-scale hierarchical fusion shift modulation weight vector of digital commodity orders, V 3d T represents the transposed vector of the digital commodity order shift modulation differential representation vector, d represents the length of the digital commodity order shift convolutional coding representation vector, Indicates position difference, Conv 1D represents a one-dimensional convolutional layer, [V1; V2] represents the concatenation of the multimodal association feature vector of the digital product order and the semantic understanding feature vector of the user behavior log, Relu represents a linear rectification function, represents matrix multiplication, Softmax represents the normalized exponential function, ⊙ represents the positional point multiplication, and V1' represents the optimized multimodal association feature vector of digital product orders.

[0049] In the technical solution of the present application, in the process of fusing the multimodal association feature vector of the digital commodity order and the semantic understanding feature vector of the user behavior log, the fine-grained inconsistency of the feature distribution between the multimodal association feature vector of the digital commodity order and the semantic understanding feature vector of the user behavior log is taken into consideration, which will cause the fused product recommendation list report generation feature vector to have abnormal stacking on the boundary of the feature manifold. Based on this, in the technical solution of the present application, before fusing the multimodal association feature vector of the digital commodity order and the semantic understanding feature vector of the user behavior log to obtain the product recommendation list report generation feature vector, the multimodal association feature vector of the digital commodity order is subjected to linear signal modulation based on fine-grained features based on the semantic understanding feature vector of the user behavior log.

[0050] In the technical solution of the present application, a linear signal modulation method based on fine-grained features is proposed to address the problem of fine-grained inconsistency in feature distribution that occurs in the process of fusing multimodal association feature vectors of digital product orders and semantic understanding feature vectors of user behavior logs. The method aims to reduce the abnormal stacking phenomenon of feature vectors generated by product recommendation list reports on the boundary of feature manifolds, so as to improve the robustness and classification performance of the model.

[0051] Specifically, the method first calculates the element-by-element difference between the digital commodity order multimodal association feature vector and the user behavior log semantic understanding feature vector to obtain the digital commodity order shift element-by-element differential representation vector, thereby capturing the difference between the two feature vectors. Then, the digital commodity order multimodal association feature vector and the user behavior log semantic understanding feature vector are cascaded into a digital commodity order joint feature vector and input into a one-dimensional convolution layer to generate a digital commodity order shift convolution coding representation vector. That is, the deep-level shift implicit information between the digital commodity order multimodal association feature vector and the user behavior log semantic understanding feature vector is captured by one-dimensional convolution coding. At the same time, the digital commodity order multimodal association feature vector is subjected to digital commodity order linear embedding modulation coding using the first digital commodity order linear embedding weight matrix to obtain a digital commodity order multimodal association feature vector linear modulation vector, and the element-by-element difference between the digital commodity order multimodal association feature vector linear modulation vector and the user behavior log semantic understanding feature vector feature vector is calculated to obtain a digital commodity order shift modulation differential representation vector. The digital commodity order shift modulation differential representation vector is used to represent the positional expression of the shift information of the digital commodity order multimodal association feature vector and the user behavior log semantic understanding feature vector in the linear space. That is, in the technical solution of the present application, the digital commodity order shift position difference, the digital commodity order shift convolutional coding representation vector and the digital commodity order shift modulation differential representation vector respectively represent the explicit expression of the fine-grained distribution difference shift information between the digital commodity order multimodal association feature vector and the user behavior log semantic understanding feature vector in different feature spaces.

[0052] Next, the shifted element-by-element differential representation vector of the digital product order is used as the value vector, the shifted convolutional coding representation vector of the digital product order and the shifted modulation differential representation vector of the digital product order are used as the query vector and the key vector respectively, and these vectors are input into the shift corrector based on the simulated transformer structure to generate the multi-scale hierarchical fusion shift modulation weight vector of the digital product order. In this way, the importance of features of different scales is evaluated by the attention mechanism, that is, the attention mechanism is used to simulate the hierarchical precipitation expression of the fine-grained difference feature distribution between the multi-modal association feature vector of the digital product order and the semantic understanding feature vector of the user behavior log in different feature spaces. Finally, the element-by-element product of the multi-scale hierarchical fusion shift modulation weight vector of the digital product order and the multi-modal association feature vector of the digital product order is calculated to obtain the optimized multi-modal association feature vector of the digital product order, that is, the weighted operation is used to assign weights to each element in the multi-modal association feature vector of the digital product order, thereby reducing the influence of local disturbances and improving the quality of the feature vector.

[0053] In summary, in the method, the digital commodity order shift position difference, the digital commodity order shift convolutional coding representation vector and the digital commodity order shift modulation difference representation vector respectively represent the explicit expression of the fine-grained distribution difference shift information between the digital commodity order multimodal association feature vector and the user behavior log semantic understanding feature vector in different feature spaces. The generation of the multi-scale hierarchical fusion shift modulation weight vector of the digital commodity order uses the attention mechanism to evaluate the importance of features at different scales and provide weight information for feature shift correction. Ultimately, the generation of the optimized digital commodity order multimodal association feature vector reduces the impact of disturbances, improves the model's robustness to local disturbances in the input data, enhances the model's adaptability to noisy or variant data sets, improves the fusion quality of the feature vector, and improves the overall classification performance and training efficiency of the model.

[0054] In a specific embodiment of the present application, the user behavior log feature extraction unit 124 includes: segmenting the user behavior log to obtain multiple user behavior log data items; passing the multiple user behavior log data items through a user behavior log context encoder including an embedding layer to obtain the user behavior log semantic understanding feature vector.

[0055] It should be understood that user behavior logs usually contain various records of users' activities in the system, such as clicks, browsing, searching, purchasing, etc. These log data are usually unstructured. In order to extract useful information from these data, word segmentation is a key step. The process of word segmentation is actually to divide the continuous text or data stream in the user behavior log into the smallest unit (i.e., data item) with practical meaning. In natural language processing, word segmentation can help to divide long text into vocabulary units for better analysis of text content. Similarly, when processing user behavior logs, word segmentation breaks down complex data in the log into specific behavior items, such as the specific page clicked by the user, the time period of the visit, the type of operation, etc. These data items can be regarded as separate entities, which helps to extract and analyze user behavior patterns. Through word segmentation, the user behavior log can be converted into multiple specific behavior data items, so that these data items can be analyzed independently. The data items after word segmentation can more clearly show the user's behavior trajectory and activity pattern, providing a basis for subsequent data processing and analysis.

[0056] Furthermore, user behavior log data items usually exist in raw, discrete forms. Although these data items provide the specific behavior of users, they do not directly reveal the deep meaning and patterns behind the behavior. In order to extract useful semantic information from these discrete behavior data, they need to be converted into more semantic feature representations. The embedding layer is one of the core tools to achieve this conversion. The embedding layer represents the data items as semantic vectors by mapping the discrete behavior data items into a continuous, high-dimensional vector space. These vectors capture the relationship and similarity between the behavior data items in the high-dimensional space, allowing the model to understand and process the intrinsic meaning of these data items. The user behavior log context encoder further processes these feature vectors based on the embedding layer, considering the contextual relationship and time series dependency between the data items. The context encoder integrates these embedding vectors through deep learning techniques (such as recurrent neural networks (RNNs) or self-attention mechanisms) to generate a contextual feature representation that fully reflects the user behavior. This process can capture the dynamic changes and contextual information of user behavior. Finally, the user behavior log semantic understanding feature vector processed by the context encoder not only integrates the embedding information of each behavior data item, but also incorporates the contextual relationship of the behavior. Specifically, the embedding layer of the user behavior log context encoder including the embedding layer is used to map each user behavior word in the user behavior word sequence into a word embedding vector to obtain a sequence of user behavior word embedding vectors; the converter-based Bert model of the user behavior log context encoder including the embedding layer is used to perform global context semantic encoding on the sequence of user behavior word embedding vectors to obtain multiple user behavior feature vectors; and the multiple user behavior feature vectors are cascaded to obtain the user behavior log semantic understanding feature vector.

[0057] In the above-mentioned digital commodity order data analysis system 100 based on big data, the associated commodity recommendation report generation module 130 is used to generate a personalized user purchase associated analysis commodity recommendation list report based on the optimized digital commodity order multimodal associated feature vector and the user behavior log semantic understanding feature vector. It should be understood that the optimized digital commodity order multimodal associated feature vector integrates the characteristics of multiple data sources, including purchase channels and order content, and can comprehensively reflect the user's behavior patterns and specific information of orders in different purchase environments. On the other hand, the user behavior log semantic understanding feature vector captures the user's interests, preferences and behavior patterns through in-depth analysis of user behavior data. Combining these two types of feature vectors, personalized recommendations for users can be achieved. By comprehensively analyzing the multimodal features of digital commodity orders and the semantic understanding of user behavior, the system can identify the user's potential interest in certain commodities or commodity categories, and generate a personalized recommendation list based on these interests. Based on this information, the system can recommend commodities that match the user's purchase history and interests, thereby improving the relevance and effectiveness of the recommendation.

[0058] Figure 3 FIG. 1 is a block diagram of a module for generating a recommendation report for related products in a digital product order data analysis system based on big data according to an embodiment of the present application. Figure 3 As shown, in a specific embodiment of the present application, the associated product recommendation report generation module 130 includes: a digital product feature fusion unit 131, which is used to fuse the optimized digital product order multimodal association feature vector and the user behavior log semantic understanding feature vector to obtain a product recommendation list report generation feature vector; a personalized associated product recommendation list report generation unit 132, which is used to pass the product recommendation list report generation feature vector through a generator to generate a personalized user purchase association analysis product recommendation list report.

[0059] It should be understood that the fusion of these two types of feature vectors can combine the specific characteristics of the product with the user's behavior pattern to generate a comprehensive recommendation feature vector. This fusion can not only improve the understanding of user preferences, but also enhance the accuracy of recommendations.

[0060] Furthermore, the feature vectors generated by the product recommendation list report are usually highly structured data representations, which contain multimodal information of the product and the behavioral characteristics of the user. Although the feature vectors generated by the product recommendation list report can accurately represent the user's preferences and the attributes of the product, they do not directly provide users with actionable recommendation information. Therefore, it is necessary to use a generator to convert the feature vectors generated by the product recommendation list report into a specific product recommendation list. By optimizing the display of recommended content, the generator can improve the user's experience and satisfaction, thereby converting highly structured feature vectors into actual recommended content.

[0061] In summary, the embodiment of the present application first obtains the digital product order purchase channel information, digital product order data and user behavior log collected by the database, and then uses deep learning technology to perform feature extraction and association analysis on the three, and finally generates a personalized user purchase association analysis product recommendation list report through a generator, thereby improving the relevance and attractiveness of the products, deeply matching the user resource upper limit, and then promoting business growth and user satisfaction.

[0062] As described above, the digital commodity order data analysis system 100 based on big data according to the embodiment of the present application can be implemented in various terminal devices. In one example, the digital commodity order data analysis system 100 based on big data can be integrated into the terminal device as a software module and / or a hardware module. For example, the digital commodity order data analysis system 100 based on big data 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 digital commodity order data analysis system 100 based on big data can also be one of the many hardware modules of the terminal device.

[0063] Alternatively, in another example, the big data-based digital commodity order data analysis system 100 and the terminal device may also be separate devices, and the big data-based digital commodity order data analysis system 100 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.

[0064] Figure 4 FIG. 1 is a flow chart of a method for analyzing digital commodity order data based on big data according to an embodiment of the present application. Figure 4As shown, according to the big data-based digital commodity order data analysis method of the embodiment of the present application, it includes: S110, obtaining digital commodity order purchase channel information, digital commodity order data and user behavior logs collected by a database; S120, extracting optimized digital commodity order multimodal association feature vectors and user behavior log semantic understanding feature vectors from the digital commodity order purchase channel information collected by the database, the digital commodity order data and the user behavior log; S130, generating a personalized user purchase association analysis product recommendation list report based on the optimized digital commodity order multimodal association feature vectors and the user behavior log semantic understanding feature vectors.

[0065] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned digital commodity order data analysis method based on big data have been referred to above. Figures 1 to 3 The invention has been introduced in detail in the description of the digital commodity order data analysis system based on big data, and therefore, its repeated description will be omitted.

[0066] Below, reference Figure 5 To describe an electronic device according to an embodiment of the present application.

[0067] See also Figure 5 , Figure 5 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0068] The processor 11 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solution provided in the embodiment of the present application; the memory 12 can be implemented in the form of a read-only memory (Read Only Memory, ROM), a static storage device, a dynamic storage device or a random access memory (Random Access Memory, RAM), etc.

[0069] The memory 12 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 12, and the processor 11 calls and executes the digital commodity order data analysis method based on big data in the embodiment of this application;

[0070] Input / output interface 13, used to implement information input and output;

[0071] The communication interface 14 is used to realize the communication interaction between the device and other devices. The communication can be realized through a wired manner (such as USB, network cable, etc.) or a wireless manner (such as mobile network, WIFI, Bluetooth, etc.);

[0072] A bus 15 that transmits information between the various components of the device (e.g., the processor 11, the memory 12, the input / output interface 13, and the communication interface 14);

[0073] The processor 11 , the memory 12 , the input / output interface 13 and the communication interface 14 are connected to each other in communication within the device via a bus 15 .

[0074] An embodiment of the present application also provides a storage medium, which is a computer-readable storage medium. The storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned digital commodity order data analysis method based on big data is implemented.

[0075] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0076] The embodiments 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. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0077] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0078] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, 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.

[0079] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0080] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0081] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0082] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods 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 through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0083] The units described above 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.

[0084] In addition, each functional unit in each embodiment of the present application can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit. If the integrated unit 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 application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method of each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs. The preferred embodiment of the embodiment of the present application is described above with reference to the accompanying drawings, and the scope of rights of the embodiment of the present application is not limited thereto. Any modifications, equivalent substitutions and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of the present application shall be within the scope of rights of the embodiments of the present application.

Claims

1. A digital commodity order data analysis system based on big data, characterized in that: include: A digital commodity order data acquisition module is used to acquire digital commodity order purchase channel information, digital commodity order data and user behavior logs collected from the database; A digital commodity order data extraction module, used to extract an optimized digital commodity order multimodal association feature vector and a user behavior log semantic understanding feature vector from the digital commodity order purchase channel information collected by the database, the digital commodity order data and the user behavior log; The associated product recommendation report generation module is used to generate a personalized user purchase association analysis product recommendation list report based on the optimized digital product order multimodal association feature vector and the user behavior log semantic understanding feature vector.

2. The digital commodity order data analysis system based on big data according to claim 1 is characterized in that: The digital commodity order data extraction module includes: A purchase channel information feature extraction unit, used to extract features from the digital commodity order purchase channel information collected from the database to obtain a digital commodity order purchase channel semantic feature vector; A commodity order data feature extraction unit, used to extract features from the digital commodity order data to obtain a digital commodity order semantic association feature vector; A commodity order multimodal feature aggregation unit, used for performing feature aggregation on the digital commodity order purchase channel semantic feature vector and the digital commodity order semantic association feature vector to obtain the optimized digital commodity order multimodal association feature vector; The user behavior log feature extraction unit is used to extract features from the user behavior log to obtain a semantic understanding feature vector of the user behavior log.

3. The digital commodity order data analysis system based on big data according to claim 2 is characterized in that: The purchase channel information feature extraction unit comprises: Passing the digital commodity order purchase channel information collected from the database through an order purchase channel semantic encoder based on a bidirectional long short-term memory neural network to obtain a plurality of digital commodity order purchase channel input vectors; The multiple digital commodity order purchase channel input vectors are passed through a digital commodity order purchase channel information semantic feature extractor based on a convolutional neural network to obtain the digital commodity order purchase channel semantic feature vector.

4. The digital commodity order data analysis system based on big data according to claim 3 is characterized in that: The commodity order data feature extraction unit comprises: Passing the digital commodity order data through a converter-based digital commodity order context encoder to obtain a plurality of digital commodity order feature vectors; The multiple digital commodity order feature vectors are concatenated into the digital commodity order semantic association feature vector.

5. The digital commodity order data analysis system based on big data according to claim 4 is characterized in that: The commodity order multimodal feature aggregation unit includes: Associating the digital commodity order purchase channel semantic feature vector with the digital commodity order semantic association feature vector to obtain a digital commodity order multimodal association feature matrix; Passing the digital commodity order multimodal association feature matrix through a digital commodity order multimodal feature encoder based on a deep convolutional neural network to obtain a digital commodity order multimodal association feature vector; Based on the semantic understanding feature vector of the user behavior log, the digital commodity order multimodal association feature vector is subjected to linear signal modulation based on fine-grained features to obtain the optimized digital commodity order multimodal association feature vector.

6. The digital commodity order data analysis system based on big data according to claim 5 is characterized in that: Based on the semantic understanding feature vector of the user behavior log, linear signal modulation based on fine-grained features is performed on the multimodal correlation feature vector of the digital commodity order to obtain the optimized multimodal correlation feature vector of the digital commodity order, including: Calculating the position difference between the digital commodity order multimodal association feature vector and the user behavior log semantic understanding feature vector to obtain a digital commodity order shift position difference representation vector; After concatenating the digital commodity order multimodal association feature vector and the user behavior log semantic understanding feature vector into a digital commodity order joint feature vector, the digital commodity order joint feature vector is input into a one-dimensional convolutional layer to obtain a digital commodity order shift convolutional coding representation vector; After performing digital commodity order linear embedding modulation encoding on the digital commodity order multimodal association feature vector using the first digital commodity order linear embedding weight matrix to obtain a digital commodity order multimodal association feature vector linear modulation vector, a position difference between the digital commodity order multimodal association feature vector linear modulation vector and the user behavior log semantic understanding feature vector is calculated to obtain a digital commodity order shift modulation difference representation vector; Taking the digital commodity order shifted position differential representation vector as the value vector, the digital commodity order shifted convolutional coding representation vector and the digital commodity order shifted modulation differential representation vector as the query vector and key vector, the digital commodity order shifted position differential representation vector, the digital commodity order shifted convolutional coding representation vector and the digital commodity order shifted modulation differential representation vector are input into a shift corrector based on an imitation transformer structure to obtain a multi-scale hierarchical fusion shift modulation weight vector for the digital commodity order; The optimized digital commodity order multimodal association feature vector is obtained by calculating the point-by-point multiplication between the multi-scale hierarchical fusion shift modulation weight vector of the digital commodity order and the multimodal association feature vector of the digital commodity order.

7. The digital commodity order data analysis system based on big data according to claim 6 is characterized in that: The user behavior log feature extraction unit includes: Segmenting the user behavior log to obtain a plurality of user behavior log data items; The multiple user behavior log data items are passed through a user behavior log context encoder including an embedding layer to obtain the user behavior log semantic understanding feature vector.

8. The digital commodity order data analysis system based on big data according to claim 7 is characterized in that: The associated product recommendation report generating module includes: A digital commodity feature fusion unit, used to fuse the optimized digital commodity order multimodal association feature vector and the user behavior log semantic understanding feature vector to obtain a commodity recommendation list report generation feature vector; The personalized associated commodity recommendation list report generating unit is used to generate a feature vector from the commodity recommendation list report through a generator to generate a personalized user purchase association analysis commodity recommendation list report.

9. A digital commodity order data analysis method based on big data, characterized in that: include: Obtain digital product order purchase channel information, digital product order data and user behavior logs collected from the database; Extracting an optimized digital commodity order multimodal association feature vector and a user behavior log semantic understanding feature vector from the digital commodity order purchase channel information collected by the database, the digital commodity order data and the user behavior log; Based on the optimized digital commodity order multimodal association feature vector and the user behavior log semantic understanding feature vector, a personalized user purchase association analysis commodity recommendation list report is generated.

10. The method for analyzing digital commodity order data based on big data according to claim 9, characterized in that: Extracting an optimized digital commodity order multimodal association feature vector and a user behavior log semantic understanding feature vector from the digital commodity order purchase channel information collected by the database, the digital commodity order data, and the user behavior log includes: Extracting features from the digital commodity order purchase channel information collected from the database to obtain a digital commodity order purchase channel semantic feature vector; Extracting features from the digital commodity order data to obtain a digital commodity order semantic association feature vector; Performing feature aggregation on the digital commodity order purchase channel semantic feature vector and the digital commodity order semantic association feature vector to obtain the optimized digital commodity order multimodal association feature vector; Feature extraction is performed on the user behavior log to obtain a semantic understanding feature vector of the user behavior log.

Images