Interest recommendation algorithm combined with user behavior data

Abstract

The invention belongs to the technical field of electronic commerce, and particularly relates to an interest recommendation algorithm combined with user behavior data, which comprises the following steps: collecting user behavior data including click data and commodities clicked by a user; browsing history, commodity categories browsed by the user and specific commodities; data processing is carried out, different weights are given to different behaviors through collected behavior data, the weight of the purchasing behavior is larger than that of the clicking behavior, and the weight of the behavior of adding the shopping cart is larger than that of pure browsing; feature extraction: extracting important features in the behavior data of the user, and extracting preference features of the user according to the behavior data of the user; according to the recommendation algorithm, after behavior data and commodity features are combined, the algorithm is adopted for recommendation, recommendation based on collaborative filtering and collaborative filtering between users, the shopping efficiency and the user experience are improved through deep learning and personalized customization, the operation strategy of the e-commerce platform is optimized, and the user experience is improved. And the effect of dual upgrading of business and user experience is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of e-commerce, and in particular to an interest recommendation algorithm combining user behavior data. Background Art

[0002] With the rapid development of mobile Internet, e-commerce mini programs, as a new e-commerce model, have become one of the main ways for users to obtain goods and services due to their lightweight and convenient features. In order to improve user experience and increase conversion rates, merchants urgently need to accurately recommend products that meet user interests and enhance user stickiness and activity.

[0003] Traditional e-commerce platforms often rely on simple recommendation methods such as product classification and keyword matching, which makes it difficult to accurately capture users' detailed interests and needs. To overcome this limitation, in recent years, scholars have been constantly exploring artificial intelligence recommendation algorithms based on user behavior data. For example, recommendation algorithms based on collaborative filtering, content filtering, and deep learning have been widely used in e-commerce platforms and have achieved remarkable results.

[0004] However, most existing recommendation algorithms are targeted at large e-commerce platforms, lacking targeted optimization that takes into account the user behavior characteristics and insufficient data resources in the specific scenarios of e-commerce mini-programs. For the user value acquisition scenarios of e-commerce mini-programs, it is urgent to develop an interest recommendation algorithm that can efficiently utilize user behavior data and comprehensively consider multiple factors such as user preferences, historical behaviors, and product characteristics, so as to achieve accurate recommendations, improve user satisfaction, and increase merchant revenue;

[0005] In response to the above technical defects, this paper proposes an interest recommendation algorithm solution that combines user behavior data. Summary of the invention

[0006] To solve the above problems, the present invention provides the following technical solutions:

[0007] An interest recommendation algorithm that combines user behavior data, including:

[0008] Data collection: collects user behavior data, including click data, the products clicked by users; browsing history, the product categories and specific products browsed by users; purchase history, the products actually purchased by users; search history, the keywords entered by users in the search box;

[0009] Data processing: different behaviors are given different weights through the collected behavioral data. The weight of purchase behavior is greater than that of click behavior, and the weight of adding to the shopping cart behavior is greater than that of simple browsing. Analysis of user interests will change over time, and the time decay model is used to weight the behavioral data.

[0010] Feature extraction: extract important features from user behavior data. Based on user behavior data, extract user preference features. Based on product attributes, extract key features of products. Combine time and location factors to extract contextual features of user behavior.

[0011] The recommendation algorithm uses an algorithm to make recommendations after combining behavioral data and product features. Recommendations based on collaborative filtering: user-to-user collaborative filtering makes recommendations based on the preferences of other users who have similar behaviors to the current user. Item-to-item collaborative filtering recommends other products similar to the products browsed and purchased by the user. Products with similar attributes are recommended based on the user's past behavior and preferences. Recommendations based on deep learning use deep learning models to learn users' potential preference patterns from large amounts of behavioral data, make personalized recommendations, build a graph structure between users and products, use graph neural networks to make recommendations, generate user portraits through behavioral data, and make recommendations based on user portraits and common features of user groups.

[0012] Furthermore, the data collection includes recording the specific time of user clicks, analyzing the user's active time periods and behavioral habits, recording the products that the user has browsed, including the browsing time and stay time, the product categories that the user has browsed, reflecting the user's interest areas, the modifications made by the user to the products in the shopping cart, and the product information actually purchased by the user, including product ID, quantity, and price; the time when the user purchased the product, which is used to analyze the purchase cycle and trends, the keywords entered by the user in the search box, reflecting the user's current needs and interests, the search results clicked by the user, to help understand the user's interest in the search results, the product information collected by the user, including product ID and collection time, the user's evaluation and rating of the product, providing subjective feedback on the product, including age, gender, geographic location, etc. = basic information, for user group segmentation and personalized recommendations, the time when user behavior occurs to analyze the time pattern of behavior, the user's sharing of products to friends or on social platforms, and the user's clicks on and participation in promotional activities and discount information.

[0013] Furthermore, the data processing includes checking and removing duplicate records to ensure the uniqueness of the data, filling in missing data, commonly used methods include mean filling, median filling, interpolation, or deleting data rows with missing values, identifying and correcting errors in the data, including format errors and inconsistent values, merging data from different sources into a unified data set, ensuring that the fields and formats in different data sources are consistent, and facilitating subsequent analysis and modeling.

[0014] Furthermore, the data processing also includes extracting features from the original data, recording the frequency of users clicking on a certain product, recording the time users stay on the product page, recording the number of times users purchase a certain type of product, extracting users' preferences for different categories of products by analyzing users' click and purchase records, recording the time characteristics of user behaviors through contextual features, recording users' geographic locations, cities or regions, scaling the data to a specific range to eliminate the impact of different feature scales, converting the data into a standard normal distribution with a mean of 0 and a standard deviation of 1 so that the data has the same weight at different scales, selecting features that have a significant impact on model predictions, removing redundant or irrelevant features, reducing high-dimensional data to a lower dimension, improving the computational efficiency of the model, and reducing the risk of overfitting, dividing the data set into a training set and a test set, the training set is used to train the model, and the test set is used to evaluate the performance of the model, dividing the data into a training set, a validation set and a test set, the validation set is used to tune model parameters, and the test set is used to evaluate the performance of the model.

[0015] Furthermore, the data processing includes normalizing the data set and scaling the data in the data set. The normalization formula is as follows:

[0016]

[0017] Among them, x' is the normalized data, x is the original data, max(X) and min(X) are the maximum and minimum values ​​of the original data.

[0018] Furthermore, the data processing includes determining the correlation of data pairs in the data set, and calculating the correlation of two associated data nodes in the same data set at the same time node. The formula is as follows:

[0019]

[0020] Among them, τ is the correlation coefficient, i is the data node in the data set, n is the number of data pairs in the data set, and x i and i is the observed value of the associated data pair in the data set, and is the average value of the associated data pairs in the dataset.

[0021] Furthermore, the feature extraction includes data cleaning, processing missing values, outliers and erroneous data, scaling the data to the same range to eliminate the scale effects of different features, identifying and retaining features that are important to the model through feature selection, evaluating the correlation between features and target variables, selecting features with larger variances, which usually contain more information, using feature importance algorithms to score the importance of features, creating new features or combining existing features to enhance the expressiveness of the model, performing sliding window aggregation on time series data, calculating mean, maximum and minimum statistical features, calculating basic statistics or using advanced methods to extract frequency domain features, using edge detection, texture analysis and convolutional neural network methods to extract image features, calculating autocorrelation, sliding window statistics and Fourier transform time series features, reducing the complexity of the feature space by selecting the most relevant features and dimensionality reduction techniques, selecting features that have a significant impact on model predictions, removing redundant or irrelevant features, and using principal component analysis and linear discriminant analysis methods to reduce data to lower dimensions to reduce computational effort and overfitting risks.

[0022] Furthermore, the feature extraction includes information gain, which is used to select the optimal partition attribute in the decision tree model, and its formula is:

[0023] InfoGain(A)=Info(D)-Info(D|A),

[0024] Among them, InfoGain(A) is the information gain of the given feature A, Info(D) is the entropy of the selected data set D, and Info(D|A) is the conditional entropy of the given feature A.

[0025] Furthermore, the feature extraction includes linear discriminant analysis, which extracts features by maximizing the inter-class divergence and minimizing the intra-class divergence. The calculation formula is as follows:

[0026]

[0027] Among them, W is the weight matrix of the principal component, det(S B ) is the inter-class scatter matrix, det(S W ) is the intra-class scatter matrix.

[0028] Furthermore, the recommendation algorithm includes calculating feature vectors of items, recommending similar items based on user interest in item features, calculating TF-IDF values ​​of items, using word embedding to represent text features, using neural network models to model complex relationships between users and items, selecting appropriate recommendation algorithms and models based on demand, training the selected recommendation model using training data, optimizing model parameters, evaluating model performance using evaluation indicators, ensuring the generalization ability and stability of the model through cross-validation methods, generating a recommendation list for the user based on the trained model, sorting and filtering the recommendation results to ensure the relevance and diversity of the recommendation results, deploying the trained recommendation model to a production environment for user use, monitoring the operation of the recommendation system, collecting user feedback and new data, regularly updating the model and recommendation algorithm to adapt to changes in user behavior and preferences, obtaining user feedback on recommendation results, analyzing user satisfaction and preference changes, optimizing the recommendation algorithm and model based on user feedback and new data, and improving the effectiveness of the recommendation system.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] In an interest recommendation algorithm combining user behavior data of the present invention, the user's behavior data is collected, including click data, that is, the products clicked by the user; browsing history, the product categories and specific products browsed by the user; purchase history, the products actually purchased by the user; search history, the keywords entered by the user in the search box; different weights are given to different behaviors through the collected behavior data, the purchase behavior weight is greater than the click behavior, and the behavior of adding to the shopping cart has a greater weight than the simple browsing behavior; the user's interest will change with time, and the time decay model is used to weight the behavior data to extract important features in the user's behavior data, and the user's preference features are extracted according to the user's behavior data, and the key features of the product are extracted according to the attributes of the product; the context features of the user's behavior are extracted by combining time and place factors; after combining the behavior data and the product features Finally, algorithms are used for recommendation. For recommendations based on collaborative filtering, users and users collaborate to filter, and make recommendations based on the preferences of other users who have similar behaviors to the current user. For collaborative filtering of items and items, other items similar to the items are recommended based on the items browsed and purchased by the user. Products with similar attributes are recommended based on the user's past behaviors and preferences. For recommendations based on deep learning, deep learning models are used to learn users' potential preference patterns from a large amount of behavioral data, and personalized recommendations are made. A graph structure between users and products is constructed, and graph neural networks are used for recommendation. User portraits are generated through behavioral data, and recommendations are made based on user portraits and the common characteristics of user groups. Deep learning and personalized customization have improved shopping efficiency and user experience, optimized the operating strategy of the e-commerce platform, and achieved the dual upgrade effect of business and user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to facilitate understanding by those skilled in the art, the present invention is further described below in conjunction with the accompanying drawings;

[0032] Figure 1 This is a schematic diagram of the overall structure of an interest recommendation algorithm that combines user behavior data according to the present invention;

[0033] Figure 2 A schematic diagram of the structure of a computer device in an interest recommendation algorithm combining user behavior data according to the present invention. DETAILED DESCRIPTION

[0034] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0035] like Figure 1-2 As shown, an interest recommendation algorithm combining user behavior data includes:

[0036] S1: Data collection, collects user behavior data, including click data, the products clicked by users; browsing history, the product categories and specific products browsed by users; purchase history, the products actually purchased by users; search history, the keywords entered by users in the search box;

[0037] S2: Data processing: different behaviors are given different weights through the collected behavioral data. The weight of purchase behavior is greater than that of click behavior, and the weight of adding to shopping cart behavior is greater than that of simple browsing. The user's interests will change over time, and the time decay model is used to weight the behavioral data.

[0038] S3: Feature extraction: extract important features from user behavior data. Based on the user behavior data, extract the user's preference features. Based on the attributes of the products, extract the key features of the products. Combine time and location factors to extract the contextual features of user behavior.

[0039] S4: Recommendation algorithm, after combining behavioral data and product features, uses an algorithm to make recommendations. Recommendations based on collaborative filtering, user-to-user collaborative filtering, make recommendations based on the preferences of other users who have similar behaviors to the current user; item-to-item collaborative filtering, recommend other products similar to the product based on the products browsed and purchased by the user, and recommend products with similar attributes based on the user's past behavior and preferences; recommendations based on deep learning, through deep learning models, learn the user's potential preference patterns from a large amount of behavioral data, make personalized recommendations, build a graph structure between users and products, use graph neural networks for recommendations, generate user portraits through behavioral data, and make recommendations based on user portraits and common characteristics of user groups.

[0040] Specifically, the data collection includes recording the specific time of user clicks, analyzing users' active time periods and behavioral habits, recording the products that users have browsed, including the browsing time and stay time, the product categories that users browsed, reflecting the users' areas of interest, the modifications that users made to the products in the shopping cart, and the product information actually purchased by users, including product ID, quantity, and price; the time when users purchased products, which is used to analyze purchase cycles and trends; the keywords entered by users in the search box, which reflect the users' current needs and interests; the search results clicked by users, which help understand the users' interest in the search results; the product information collected by users, including product ID and collection time; users' evaluations and ratings of products, providing subjective feedback on products, including age, gender, geographic location, etc. = basic information, for user group segmentation and personalized recommendations; the time when user behavior occurs to analyze the time pattern of behavior; users' sharing of products with friends or sharing on social platforms; and users' clicks on and participation in promotional activities and discount information.

[0041] Specifically, the data processing includes checking and removing duplicate records to ensure the uniqueness of the data, filling in missing data. Common methods include mean filling, median filling, interpolation, or deleting data rows with missing values, identifying and correcting errors in the data, including format errors and inconsistent values, merging data from different sources into a unified data set, ensuring that the fields and formats in different data sources are consistent, and facilitating subsequent analysis and modeling.

[0042] Specifically, the data processing also includes extracting features from the original data, recording the frequency of users clicking on a certain product, recording the time users stay on the product page, recording the number of times users purchase a certain type of product, extracting users' preferences for different categories of products by analyzing users' click and purchase records, recording the time characteristics of user behaviors through contextual features, recording users' geographic locations, cities or regions, scaling the data to a specific range to eliminate the impact of different feature scales, converting the data to a standard normal distribution with a mean of 0 and a standard deviation of 1 so that the data has the same weight at different scales, selecting features that have a significant impact on model predictions, removing redundant or irrelevant features, reducing high-dimensional data to a lower dimension, improving the computational efficiency of the model, and reducing the risk of overfitting, dividing the data set into a training set and a test set, the training set is used to train the model, and the test set is used to evaluate the performance of the model, dividing the data into a training set, a validation set and a test set, the validation set is used to tune model parameters, and the test set is used to evaluate the performance of the model.

[0043] Specifically, the data processing includes normalizing the data set and scaling the data in the data set. The normalization formula is as follows:

[0044]

[0045] Among them, x' is the normalized data, x is the original data, max(X) and min(X) are the maximum and minimum values ​​of the original data.

[0046] Specifically, the data processing includes judging the relevance of data pairs in the data set, and calculating the relevance of two associated data nodes in the same data set at the same time node. The formula is as follows:

[0047]

[0048] Among them, τ is the correlation coefficient, i is the data node in the data set, n is the number of data pairs in the data set, and x i and i is the observed value of the associated data pair in the data set, and is the average value of the associated data pairs in the dataset.

[0049] Specifically, the feature extraction includes data cleaning, processing missing values, outliers and erroneous data, scaling the data to the same range to eliminate the scale effects of different features, identifying and retaining features that are important to the model through feature selection, evaluating the correlation between features and target variables, selecting features with larger variances, which usually contain more information, using feature importance algorithms to score the importance of features, creating new features or combining existing features to enhance the expressiveness of the model, performing sliding window aggregation on time series data, calculating mean, maximum and minimum statistical features, calculating basic statistics or using advanced methods to extract frequency domain features, using edge detection, texture analysis and convolutional neural network methods to extract image features, calculating autocorrelation, sliding window statistics and Fourier transform time series features, reducing the complexity of the feature space by selecting the most relevant features and dimensionality reduction techniques, selecting features that have a significant impact on model predictions, removing redundant or irrelevant features, and using principal component analysis and linear discriminant analysis methods to reduce data to lower dimensions to reduce the amount of calculation and the risk of overfitting.

[0050] After feature extraction and selection, the features are further engineered:

[0051] Feature combination: Combine multiple features into a new feature.

[0052] Feature transformation: Apply logarithmic transformation, square root transformation, etc. to change the distribution form of features.

[0053] Feature encoding: Convert categorical features into numerical features, such as One-Hot Encoding or Label Encoding.

[0054] Evaluate the impact of features on model performance and ensure the effectiveness of feature engineering:

[0055] Model training and validation: Use the extracted features to train the model and evaluate the model performance through methods such as cross-validation. Feature importance analysis: Analyze the importance of features through models (such as decision trees) to determine the contribution of features to the model. Through the above steps, valuable features can be extracted from the original data to provide powerful input for the machine learning model, thereby improving the model's predictive performance and accuracy. Feature extraction is an important part of data preprocessing and modeling, and directly affects the effect of the final model.

[0056] Specifically, the feature extraction includes information gain, which is used to select the optimal partition attribute in the decision tree model, and its formula is:

[0057] InfoGain(A)=Info(D)-Info(D|A),

[0058] Among them, InfoGain(A) is the information gain of the given feature A, Info(D) is the entropy of the selected data set D, and Info(D|A) is the conditional entropy of the given feature A.

[0059] Specifically, the feature extraction includes linear discriminant analysis, which extracts features by maximizing the inter-class divergence and minimizing the intra-class divergence. The calculation formula is as follows:

[0060]

[0061] Among them, W is the weight matrix of the principal component, det(S B ) is the inter-class scatter matrix, det(S W ) is the intra-class scatter matrix.

[0062] Specifically, the recommendation algorithm includes calculating the feature vector of the item, recommending similar items based on the user's interest in the item features, calculating the TF-IDF value of the item, using word embedding to represent text features, using a neural network model to model the complex relationship between users and items, selecting a suitable recommendation algorithm and model according to the needs, using training data to train the selected recommendation model, optimizing model parameters, using evaluation indicators to evaluate model performance, ensuring the generalization ability and stability of the model through cross-validation methods, generating a recommendation list for the user based on the trained model, sorting and filtering the recommendation results to ensure the relevance and diversity of the recommendation results, deploying the trained recommendation model to the production environment for user use, monitoring the operation of the recommendation system, collecting user feedback and new data, regularly updating the model and recommendation algorithm to adapt to changes in user behavior and preferences, obtaining user feedback on the recommendation results, analyzing user satisfaction and preference changes, optimizing the recommendation algorithm and model based on user feedback and new data, and improving the effect of the recommendation system. The recommendation system is a process of continuous iteration and optimization, which needs to be continuously adjusted and improved according to actual application scenarios and user needs. The recommendation algorithm is used to measure the impact of the recommendation system on user retention. A system with good recommendation quality can usually improve user retention. Measure the contribution of the recommendation system to sales revenue. A good recommendation system can increase user purchases and total revenue. By dividing users into different groups, testing the effects of different recommendation algorithms, and comparing the performance of different algorithms, the best solution can be selected. Evaluating the effectiveness of a recommendation algorithm is not just about calculating technical indicators such as accuracy, but also about considering actual business needs and user experience. By using these indicators and methods in combination, the performance of the recommendation system can be comprehensively evaluated and a basis for optimizing the recommendation algorithm can be provided.

[0063] Those skilled in the art will understand that Figure 2The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0064] In one embodiment, a computer-readable storage medium is provided, storing a computer program, which, when executed by a processor, implements the steps in the above-mentioned interest recommendation algorithm embodiment combining user behavior data.

[0065] In one embodiment, a computer program product or computer program is provided, the computer program product or computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps in the above-mentioned interest recommendation algorithm embodiment combined with user behavior data.

[0066] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0067] The working principle of the interest recommendation algorithm combined with user behavior data of the present invention is as follows: by collecting user behavior data, including click data, the products clicked by the user; browsing history, the product categories and specific products browsed by the user; purchase history, the products actually purchased by the user; search history, the keywords entered by the user in the search box; different weights are given to different behaviors through the collected behavior data, the purchase behavior weight is greater than the click behavior, and the behavior of adding to the shopping cart has a greater weight than the simple browsing behavior; the user's interest will change with time, and the time decay model is used to weight the behavior data to extract important features in the user behavior data, and the user's preference features are extracted according to the user's behavior data, and the key features of the product are extracted according to the attributes of the product; the context features of the user behavior are extracted by combining time and place factors; after combining the behavior data and the product characteristics After the collection, algorithms are used for recommendation. For recommendations based on collaborative filtering, users perform collaborative filtering and make recommendations based on the preferences of other users who have similar behaviors to the current user. For collaborative filtering of items, other items similar to the items are recommended based on the items browsed and purchased by the user. Products with similar attributes are recommended based on the user's past behaviors and preferences. For recommendations based on deep learning, deep learning models are used to learn users' potential preference patterns from a large amount of behavioral data, and personalized recommendations are made. A graph structure between users and products is constructed, and graph neural networks are used for recommendation. User portraits are generated through behavioral data, and recommendations are made based on user portraits and the common characteristics of user groups. Deep learning and personalized customization have improved shopping efficiency and user experience, optimized the operating strategy of the e-commerce platform, and achieved the dual upgrade effect of business and user experience.

[0068] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. An interest recommendation algorithm that combines user behavior data, It is characterized in that include: Data collection: collect user behavior data, including click data and the products clicked by users; Browsing history, the categories and specific products that users browsed; Purchase history, the products that the user has actually purchased; search history, the keywords that the user entered in the search box; Data processing: different behaviors are given different weights through the collected behavioral data. The weight of purchase behavior is greater than that of click behavior, and the weight of adding to the shopping cart behavior is greater than that of simple browsing. Analysis of user interests will change over time, and the time decay model is used to weight the behavioral data. Feature extraction: extract important features from user behavior data. Based on the user behavior data, extract the user's preference features. Based on the attributes of the product, extract the key features of the product. Combine time and location factors to extract contextual features of user behavior; Recommendation algorithm, after combining behavioral data and product features, uses algorithms to make recommendations. Recommendations based on collaborative filtering, user-to-user collaborative filtering, recommendations are made based on the preferences of other users who have similar behaviors to the current user; item-to-item collaborative filtering, based on the products browsed and purchased by the user, other products similar to the product are recommended, and based on the user's past behavior and preferences, products with similar attributes are recommended; Recommendations based on deep learning use deep learning models to learn users' potential preference patterns from large amounts of behavioral data, make personalized recommendations, build a graph structure between users and products, use graph neural networks to make recommendations, generate user portraits through behavioral data, and make recommendations based on user portraits and common characteristics of user groups.

2. According to claim 1, an interest recommendation algorithm combining user behavior data, It is characterized in that The data collection includes recording the specific time of user clicks, analyzing the user's active time periods and behavioral habits, recording the products that users browsed, including the browsing time and duration of stay, the categories of products that users browsed, reflecting the user's areas of interest, the user's modifications to the products in the shopping cart, and the information of the products actually purchased by the user, including product ID, quantity, and price; The time when users purchase products is used to analyze purchase cycles and trends. The keywords entered by users in the search box reflect the users' current needs and interests. The search results clicked by users help understand the users' interest in the search results. Information about products collected by users, including product ID and collection time. Users' comments and ratings on products provide subjective feedback on products, including basic information such as age, gender, and geographic location, for user group segmentation and personalized recommendations. The time when user behavior occurs analyzes the temporal pattern of behavior. Users share products with friends or on social platforms. Users click on and participate in promotional activities and discount information.

3. According to claim 2, an interest recommendation algorithm combining user behavior data, It is characterized in that The data processing includes checking and removing duplicate records to ensure the uniqueness of the data, filling in missing data. Common methods include mean filling, median filling, interpolation, or deleting data rows with missing values, identifying and correcting errors in the data, including format errors and inconsistent values, merging data from different sources into a unified data set, ensuring that the fields and formats in different data sources are consistent, and facilitating subsequent analysis and modeling.

4. According to claim 3, an interest recommendation algorithm combining user behavior data, It is characterized in that The data processing also includes extracting features from the original data, recording the frequency of users clicking on a certain product, recording the time users stay on the product page, recording the number of times users purchase a certain type of product, extracting users' preferences for different categories of products by analyzing users' click and purchase records, recording the time characteristics of user behaviors through contextual features, recording users' geographic locations, cities or regions, scaling the data to a specific range to eliminate the impact of different feature scales, converting the data to a standard normal distribution with a mean of 0 and a standard deviation of 1 so that the data has the same weight at different scales, selecting features that have a significant impact on model predictions, removing redundant or irrelevant features, reducing high-dimensional data to a lower dimension, improving the computational efficiency of the model, and reducing the risk of overfitting, dividing the data set into a training set and a test set, the training set is used to train the model, and the test set is used to evaluate the performance of the model, dividing the data into a training set, a validation set and a test set, the validation set is used to tune model parameters, and the test set is used to evaluate the performance of the model.

5. According to claim 4, an interest recommendation algorithm combining user behavior data, It is characterized in that The data processing includes normalizing the data set and scaling the data in the data set. The normalization formula is as follows: Among them, x' is the normalized data, x is the original data, max(X) and min(X) are the maximum and minimum values ​​of the original data.

6. The interest recommendation algorithm combined with user behavior data according to claim 5, It is characterized in that The data processing includes judging the relevance of data pairs in the data set, and calculating the relevance of two related data nodes in the same data set at the same time node. The formula is as follows: Among them, τ is the correlation coefficient, i is the data node in the data set, n is the number of data pairs in the data set, and x i and i is the observed value of the associated data pair in the data set, and is the average value of the associated data pairs in the dataset.

7. The interest recommendation algorithm combined with user behavior data according to claim 6, It is characterized in that The feature extraction includes data cleaning, processing missing values, outliers and erroneous data, scaling the data to the same range to eliminate the scale effects of different features, identifying and retaining features that are important to the model through feature selection, evaluating the correlation between features and target variables, selecting features with larger variances, which usually contain more information, using feature importance algorithms to score the importance of features, creating new features or combining existing features to enhance the expressiveness of the model, performing sliding window aggregation on time series data, calculating mean, maximum and minimum statistical features, calculating basic statistics or using advanced methods to extract frequency domain features, using edge detection, texture analysis and convolutional neural network methods to extract image features, calculating autocorrelation, sliding window statistics and Fourier transform time series features, reducing the complexity of the feature space by selecting the most relevant features and dimensionality reduction techniques, selecting features that have a significant impact on model predictions, removing redundant or irrelevant features, and using principal component analysis and linear discriminant analysis methods to reduce data to lower dimensions to reduce the amount of calculation and the risk of overfitting.

8. The interest recommendation algorithm combined with user behavior data according to claim 6, It is characterized in that The feature extraction includes information gain, which is used to select the optimal partition attribute in the decision tree model. The formula is: InfoGain(A)=Info(D)-Info(D|A), Among them, InfoGain(A) is the information gain of the given feature A, Info(D) is the entropy of the selected data set D, and Info(D|A) is the conditional entropy of the given feature A.

9. The interest recommendation algorithm combined with user behavior data according to claim 8, characterized in that the feature extraction includes linear discriminant analysis, and the features are extracted by maximizing the inter-class divergence and minimizing the intra-class divergence, and the calculation formula is as follows; in, W is the weight matrix of the principal components, det(S B ) is the inter-class scatter matrix, det(S W ) is the intra-class scatter matrix.

10. The interest recommendation algorithm combined with user behavior data according to claim 8, It is characterized in that The recommendation algorithm includes calculating feature vectors of items, recommending similar items based on user interest in item features, calculating TF-IDF values ​​of items, using word embedding to represent text features, using neural network models to model complex relationships between users and items, selecting appropriate recommendation algorithms and models based on demand, training the selected recommendation model using training data, optimizing model parameters, evaluating model performance using evaluation indicators, ensuring the generalization ability and stability of the model through cross-validation methods, generating a recommendation list for the user based on the trained model, sorting and filtering the recommendation results to ensure the relevance and diversity of the recommendation results, deploying the trained recommendation model to a production environment for user use, monitoring the operation of the recommendation system, collecting user feedback and new data, regularly updating the model and recommendation algorithm to adapt to changes in user behavior and preferences, obtaining user feedback on recommendation results, analyzing user satisfaction and preference changes, optimizing the recommendation algorithm and model based on user feedback and new data, and improving the effectiveness of the recommendation system.

Images