A user product preference analysis method and system based on data security

Abstract

The application discloses a kind of based on data security's user product preference analysis method and system, including the method includes the following steps: S1, obtain the browsing information and purchase information of user, and analysis obtains user behavior parameter;S2, the data cleaning of obtained user behavior parameter is carried out, and based on the type of product, the browsing and purchase data are classified, and classification behavior data are obtained;S3, according to classification behavior data, the subjective preference degree and objective preference degree of user to product are calculated;S4, according to pre-set reliability, the subjective preference degree and objective preference degree of user are merged, and the comprehensive preference degree of user to product is calculated;S5, according to comprehensive preference degree, the purchase preference of user to product is analyzed, and based on the comprehensive preference of user, corresponding product is recommended for user.The application carries out classification to user product preference value interpretation result and carries out desensitization processing to data, improves the use efficiency of the based on data security's user product preference analysis method.

Description

Technical Field

[0001] This invention relates to the fields of data analysis and data security technology, and more specifically, to a user product preference analysis method and system based on data security. Background Technology

[0002] With the rapid development of the Internet, people's use of the network has become increasingly widespread, and online shopping has attracted particular attention. The number of servers connected to the Internet and the number of web pages on the Internet have shown an exponential growth trend. The rapid development of Internet technology has brought a massive amount of information to our attention at the same time. The digital revolution has led to a crazy increase in the total amount of information globally. Individuals have too much information to process, and the information explosion has actually reduced the utilization rate of information. Among these factors, user product preferences have received increasing attention. Purchase preference analysis involves mining historical behavioral data and making reliable calculations to predict users' future purchasing behavior. User product preference analysis is a key market research task. It aims to understand consumers' needs and preferences for different products. At the same time, protecting user data security is very important when conducting user product preference analysis.

[0003] Data security is receiving increasing attention, and its significance lies in protecting the sensitive information of individuals and organizations from unauthorized access, use, modification, disclosure, or damage. Ensuring data security is crucial for protecting the financial, intellectual property, customer records, health records, and other sensitive information of businesses, institutions, organizations, and individuals. Effective data security helps maintain the reputation and credibility of businesses, organizations, and individuals, promotes business growth and development, reduces the impact of data breaches on customers, and improves product sales reputation.

[0004] Existing user preference analysis methods, when mining data from historical browsing logs of retail products to obtain information, focus more on assessing users' personal spending power. When predicting the types and styles of products users need, they rely more on real-time browsing data. However, they cannot conduct real-time user product preference analysis based on the impact of live-streaming e-commerce and influencer marketing in modern society. As a result, existing user product preference analysis methods cannot meet the needs of merchants.

[0005] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0006] In response to the problems in related technologies, this invention proposes a user product preference analysis method and system based on data security to overcome the aforementioned technical problems existing in the existing related technologies.

[0007] Therefore, the specific technical solution adopted by the present invention is as follows:

[0008] A user product preference analysis method based on data security, the method includes the following steps:

[0009] S1. Obtain user browsing and purchase information, and analyze it to obtain user behavior parameters;

[0010] S2. Clean the obtained user behavior parameters and classify the browsing and purchase data according to the product type to obtain classified behavior data.

[0011] S3. Calculate users' subjective and objective preferences for the product based on the classification behavior data;

[0012] S4. Combine the user's subjective and objective preferences based on the preset reliability, and calculate the user's overall preference for the product;

[0013] S5. Analyze users' purchasing preferences for products based on comprehensive preference analysis, and recommend corresponding products to users based on their comprehensive preferences.

[0014] Furthermore, the calculation of users' subjective and objective preferences for products based on classification behavior data includes the following steps:

[0015] S31. Divide the classification behavior data into subjective data and objective data, and perform anonymization processing on the subjective data and objective data;

[0016] S32. Calculate the user's objective preference based on the de-identified objective data;

[0017] S33. Calculate the user's subjective preference based on the de-identified subjective data.

[0018] Furthermore, the calculation of the user's objective preference degree based on the anonymized objective data includes the following steps:

[0019] S321. Group the desensitized objective data according to the product type to obtain objective grouped data;

[0020] S322. Establish a time series forecasting model based on the grouping results, analyze and forecast the grouped objective data based on the time series forecasting model to obtain the time series forecast values, and assign weights to the time series forecast values.

[0021] S323. Calculate the objective preference value based on the objective grouping data, and assign weights to the objective preference value;

[0022] S324. Calculate the user's objective preference for the product based on the time-series predicted value and the objective preference value after weighting.

[0023] Furthermore, the process of dividing the classification behavior data into subjective data and objective data, and then anonymizing the subjective and objective data, includes the following steps:

[0024] The anonymized subjective and objective data were categorized by region and time, and statistical analysis was performed on the anonymized subjective and objective data to obtain statistical indicators such as the mean, standard deviation, maximum and minimum values ​​of the anonymized subjective and objective data.

[0025] Furthermore, the step of establishing a time-series prediction model based on the grouping results, analyzing and predicting the grouped objective data using the time-series prediction model to obtain time-series prediction values, and assigning weights to the time-series prediction values ​​includes the following steps:

[0026] S3221. Obtain historical data based on classification behavior data, analyze the historical data to obtain objective data, and preprocess the objective data.

[0027] S3222. Construct and train an improved BP neural network model based on the preprocessed objective data;

[0028] S3223. Real-time prediction of collected objective data based on the improved BP neural network model;

[0029] S3224. Use the prediction results of the improved BP neural network model as the time-series prediction value of objective data.

[0030] Furthermore, the construction and training of the improved BP neural network model based on the preprocessed objective data includes the following steps:

[0031] S32221. Use 60% of the preprocessed objective data as the training set and 40% as the test set.

[0032] S32222. Select a personalized system recommendation algorithm to construct an improved BP neural network model, and train the model using the training set. The formula for the improved BP neural network model is:

[0033]

[0034] In the formula, RE is the time-series predicted value of objective data;

[0035] p a,u Similarity between objective data;

[0036] v u，i A rating vector for objective data;

[0037] hy a，u It is a mixed weighting factor;

[0038] wxa The number of rating values;

[0039] C v This is the first score vector for the grouped data;

[0040] S32223. Input the test machine into the improved BP neural network model after training and test it. Use cross-validation to train and test the improved BP neural network model multiple times, and calculate the average error to estimate the accuracy of the result.

[0041] S32223. Iterate the improved BP neural network model based on the estimation results.

[0042] Furthermore, the calculation of the user's subjective preference based on the anonymized subjective data includes the following steps:

[0043] S331. Conduct pre-analysis of subjective data;

[0044] S332. Calculate the subjective preference value based on the pre-analyzed subjective data, and calculate the subjective preference degree by weighting the subjective preference value.

[0045] Furthermore, the step of merging the user's subjective and objective preferences based on a preset reliability level and calculating the user's overall preference for the product includes the following steps:

[0046] S41. Assign weights to subjective and objective preferences respectively;

[0047] S42. The overall preference score is calculated based on the weights of subjective preference and objective preference.

[0048] Furthermore, the step of analyzing user purchase preferences based on comprehensive preference and recommending corresponding products to users based on their comprehensive preferences includes the following steps:

[0049] S51. Recommend similar items based on the user's overall product preference for items with a high overall preference level.

[0050] S52. Find similar items based on the user's overall product preference, increase the frequency of similar item push, and perform real-time price analysis on similar items based on the user's overall product preference, and push real-time price changes to the user.

[0051] According to another aspect of the present invention, a user product preference analysis system based on data security is provided, the system comprising:

[0052] The data acquisition module acquires users' browsing and purchase information and analyzes it to obtain user behavior parameters.

[0053] The data cleaning module is used to clean the obtained user behavior parameters and classify the browsing and purchase data based on the product type to obtain categorized behavior data.

[0054] The behavioral data analysis module is used to calculate users' subjective and objective preferences for products based on categorized behavioral data.

[0055] The comprehensive preference acquisition module is used to combine users' subjective and objective preferences based on a preset reliability, and calculate users' comprehensive preference for the product.

[0056] The recommendation adjustment module is used to analyze users' purchasing preferences for products based on their overall preference score, and recommend corresponding products to users based on their overall preferences.

[0057] The beneficial effects of this invention are as follows:

[0058] 1. This invention categorizes the interpretation results of user product preference values ​​and performs data anonymization processing, which facilitates the protection of user privacy and data security. At the same time, it facilitates real-time monitoring of product sales driven by live streaming e-commerce and influencer effects, enabling merchants to adjust product procurement in a timely manner based on user product preference analysis, greatly improving the efficiency of this data security-based user product preference analysis method.

[0059] 2. This invention improves the accuracy of user product preference analysis methods by performing time-series data analysis to observe product sales and user purchasing preferences in a timely manner. At the same time, it recommends high-selling products to users in real time by allocating weights to overall preference values ​​and individual preference values, thereby improving the analytical effect of user product preference analysis methods. Furthermore, by analyzing individual preference values, it enhances the convenience for users to purchase items that meet their needs.

[0060] 3. By finding and recommending similar items based on user product preference values, the convenience of product selection for users is improved. At the same time, recommending items that users need based on their product preference values ​​improves the sales efficiency of merchants' products. This greatly improves the efficiency of user product preference analysis methods based on data security, and provides real-time dynamic price monitoring of products that users are interested in, further enhancing the convenience of user experience. Attached Figure Description

[0061] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0062] Figure 1 This is a flowchart of a user product preference analysis method based on data security according to an embodiment of the present invention;

[0063] Figure 2 This is a system principle block diagram of a user product preference analysis system based on data security according to an embodiment of the present invention.

[0064] In the picture:

[0065] 1. Data acquisition module; 2. Data cleaning module; 3. Behavioral data analysis module; 4. Comprehensive preference acquisition module; 5. Recommendation adjustment module. Detailed Implementation

[0066] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention.

[0067] According to embodiments of the present invention, a method and system for analyzing user product preferences based on data security are provided.

[0068] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1 As shown, according to an embodiment of the present invention, a user product preference analysis method based on data security includes the following steps:

[0069] S1. Obtain user browsing and purchase information, and analyze it to obtain user behavior parameters;

[0070] S2. Clean the obtained user behavior parameters and classify the browsing and purchase data according to the product type to obtain classified behavior data.

[0071] Specifically, when classifying product data, products can be categorized broadly, such as into food, furniture, and digital products.

[0072] S3. Calculate users' subjective and objective preferences for the product based on the classification behavior data;

[0073] The calculation of users' subjective and objective preferences for products based on categorized behavioral data includes the following steps:

[0074] S31. Divide the classification behavior data into subjective data and objective data, and perform anonymization processing on the subjective data and objective data;

[0075] Specifically, data anonymization can be achieved through methods such as complete deletion, randomization, hashing, anonymization replacement, and data generalization. This solution includes the following steps for data anonymization: determining the conditions for anonymizing subjective and objective data. First, it is necessary to determine the date range, data type, and data format for anonymization, and then check the integrity of subjective and objective data to ensure there are no missing, duplicate, or erroneous entries.

[0076] Remove invalid data: Remove invalid data such as null values, blank lines, and incomplete data. Then remove duplicate data: Remove duplicate rows and duplicate records, and filter out abnormal data, such as data that exceeds the expected range and outliers. Then standardize subjective and objective data by unifying date formats and capitalization. Finally, remove personal identification information, bank account information, etc.

[0077] Finally, the data was verified using both subjective and objective data: the filtered data was verified to ensure its accuracy and completeness;

[0078] Select appropriate encryption algorithms, such as AES and RSA, for the filtered subjective and objective data, generate keys to ensure data security, and then use the selected encryption algorithm and the generated key to encrypt the user product preference data.

[0079] The process of dividing the processed behavioral data into subjective and objective data, and the data anonymization process also includes the following steps:

[0080] The desensitized subjective and objective data were categorized by region and time, and statistical analysis was performed on the desensitized subjective and objective data to obtain statistical indicators such as the mean, standard deviation, maximum and minimum values ​​of the desensitized subjective and objective data.

[0081] S32. Calculate the user's objective preference based on the de-identified objective data;

[0082] The calculation of users' objective preferences based on anonymized objective data includes the following steps:

[0083] S321. Group the desensitized objective data according to the product type to obtain objective grouped data;

[0084] Specifically, grouping is a further differentiation of products after classification. For example, products are classified into food categories, and then further classified into specific categories such as vegetarian food, meat, and soy products.

[0085] S322. Establish a time series forecasting model based on the grouping results, analyze and forecast the grouped objective data based on the time series forecasting model to obtain the time series forecast values, and assign weights to the time series forecast values.

[0086] The process of establishing a time-series forecasting model based on the grouping results, analyzing and forecasting the grouped objective data using the time-series forecasting model to obtain time-series forecast values, and assigning weights to the time-series forecast values ​​includes the following steps:

[0087] S3221. Obtain historical data based on classification behavior data, analyze the historical data to obtain objective data, and preprocess the objective data:

[0088] Data cleaning: Checking and processing outliers, missing values, duplicate values, etc. in data. Missing values ​​can be deleted or filled, outliers can be removed, or appropriate imputation methods can be used.

[0089] Data transformation: Transforming data to meet the needs of modeling, analysis, or visualization, such as data normalization, standardization, discretization, or continuousization.

[0090] Feature selection: If the data contains a large number of features or dimensions, feature selection methods can be applied to select the most relevant or informative features in order to reduce dimensionality and improve modeling performance.

[0091] Data integration: If the data comes from multiple sources or multiple tables, data integration and merging may be necessary to create a consistent dataset for subsequent analysis.

[0092] Data normalization: Mapping data to the same range or scale. Normalization methods include min-max normalization.

[0093] Data standardization: By subtracting the mean and dividing by the standard deviation, data is transformed into a standard distribution with zero mean and unit variance.

[0094] Outlier handling: Detecting and handling outliers can be done using statistical methods, rule-based methods, or machine learning methods to identify and process them.

[0095] Data resampling: If the data is imbalanced, resampling techniques, such as oversampling or undersampling, can be applied to balance the class distribution.

[0096] S3222. Construct and train an improved BP neural network model based on the preprocessed objective data;

[0097] S3223. Real-time prediction of collected objective data based on the improved BP neural network model;

[0098] S3224. Use the prediction results of the improved BP neural network model as the time-series prediction value of objective data.

[0099] The process of constructing and training an improved BP neural network model based on preprocessed objective data includes the following steps:

[0100] S32221. Use 60% of the preprocessed objective data as the training set and 40% as the test set.

[0101] S32222. Select a personalized system recommendation algorithm to construct an improved BP neural network model, and train the model using the training set. The formula for the improved BP neural network model is:

[0102]

[0103] hy a，u =hg a，u +hx a,u

[0104]

[0105]

[0106]

[0107]

[0108]

[0109] Where RE is the time-series predicted value of objective data;

[0110] p a,u The similarity between grouped data is calculated using cosine similarity. For text data or high-dimensional feature vectors, cosine similarity is used to measure the similarity between two groups.

[0111] v u，i The initial score vector is the score vector for the grouped data. The initial score vector for the grouped data is calculated using relevant data analysis methods. Depending on the data characteristics and analysis objectives, various data analysis methods can be used to generate the initial score vector. Clustering algorithms can be used to group the data and the clustering results can be used as the initial score vector.

[0112] hy a，uA mixed weighting factor is a weighting factor used to calculate a user's overall preference for a product. It assigns weights to both subjective and objective preferences, such as subjective weighting factors and objective weighting factors. Subjective weighting factors reflect the user's emphasis on personal preferences, while objective weighting factors reflect the user's emphasis on recommendations based on objective data. Mixed weighting factors are obtained by conducting surveys targeting the user group, inquiring about their emphasis on personal preferences and objective data, or by statistically analyzing user feedback data, such as calculating the mean and standard deviation, to understand the relative importance of subjective and objective factors. Based on the results of the statistical analysis, a series of experiments can be conducted to verify the effectiveness of the weighting factors. By interacting with users and observing their behavior and feedback, the weighting factors can be gradually adjusted to more accurately reflect the user's overall preference.

[0113] wx a Number of rating values: This refers to the number of rating values ​​that users give to the product. When conducting user behavior data analysis and product recommendations, rating systems are usually used to measure users' liking or satisfaction with the product. Rating systems can be discrete or continuous, and are designed according to actual needs and circumstances.

[0114] The number of rating values ​​is obtained by predefined discrete rating values. When designing a rating system, a set of discrete rating values ​​is predefined, such as a 5-star rating system, ranging from 1 star to 5 stars. Users can choose one of the rating values ​​to represent their level of liking or satisfaction with the product. The rating system can also use a continuous numerical range, such as a score from 0 to 10, or a percentage from 0 to 100. In addition to discrete rating values ​​or numerical ranges, sometimes users can be given the option of free text evaluation, allowing them to describe their evaluation of the product in their own words.

[0115] C vThe first rating vector for grouped data refers to the vector formed by the first ratings of users in each group after users are grouped. Grouping users is for better understanding and analysis of their behavior and preferences. Grouping can be based on different characteristics or criteria, such as user's geographic location, age, gender, purchase history, etc. After grouping users, the first rating of each user in each group can be calculated. The resulting data can form a vector, where each element of the vector represents a user's first rating in that group. The first rating vector for grouped data is obtained by determining the criteria used for grouping. These criteria can be based on user characteristics, such as geographic location, age, gender, purchase history, etc., selecting criteria that match the analysis objectives and needs. Using the selected grouping criteria, users are grouped according to their corresponding characteristics or attributes, dividing users into different groups, each representing a specific subgroup. For users in each group, their first rating in that group is found from their rating records. For each group, the first ratings of all users are combined into a vector.

[0116] hg a，u The first weighting factor is the weighting factor used to measure subjective preferences when calculating a user's overall preference for a product. It represents the weight of subjective preferences in the overall preference calculation, determining the importance of individual user preferences—that is, the proportion of subjective factors in a user's product preference. By adjusting the value of the first weighting factor, the influence of subjective preferences on the final preference can be adjusted; a larger first weighting factor indicates greater importance, and vice versa. The first weighting factor is determined by conducting questionnaires or user interviews, directly asking users about the importance and weight of different subjective preference factors, and collecting user opinions and feedback to understand their evaluation and relative importance of each factor.

[0117] hx a，u The second weighting factor is a weighting factor used to measure objective preferences when calculating a user's overall preference for a product. It represents the weight of objective preferences in the overall user preference calculation, determining the importance of recommendations based on objective data to user preferences—that is, the proportion of objective factors in a user's product preference. By adjusting the value of the second weighting factor, the influence of objective preferences on the final preference can be adjusted; a larger second weighting factor indicates greater importance, and vice versa. The second weighting factor is determined by observing user choices and behavioral responses in an experimental environment by showing users different products or changes in influencing factors. Based on the experimental results, the influence of different factors on the user's final preference can be analyzed, thus inferring the second weighting factor.

[0118] N a，u The number of overlapping ratings in grouped data refers to the number of users whose ratings overlap across different groups after user grouping. When users are grouped and their rating behavior is analyzed, it may be found that some users belong to multiple groups and have rating records in all of them. In such cases, the number of overlapping ratings in grouped data can be calculated, i.e., the number of users existing in multiple groups. The number of overlapping ratings in grouped data can provide further insights into user behavior and preferences. By identifying and analyzing users with overlapping ratings, consistent or similar preferences among users in different groups can be discovered. This is often of great significance for personalized recommendations and customized services. The number of overlapping ratings in grouped data is calculated by using clear group definitions and operational standards, such as geographical location, age, gender, purchase history, etc., to group users into corresponding groups, obtain user rating data, and retrieve the rating records of each user in each group. For each group, the rating records of that group are compared with those of other groups to calculate the number of users existing in multiple groups, i.e., the number of overlapping ratings.

[0119] n i The number of ratings for grouped data refers to the total number of ratings given by users in each group after user groups have been formed. When grouping users and analyzing their rating behavior, the number of ratings given by users in each group can be counted. This metric provides information about user activity and engagement within a group. The number of ratings for grouped data can be used to understand the level of user participation and engagement in different groups. A higher number of ratings indicates that users in that group are more frequent and active in rating, while a lower number of ratings indicates that users in that group are less active or less engaged in rating. By analyzing the number of ratings for grouped data, we can identify which groups of users are more likely to be interested in a product or service, and based on the ratings... The differences in the number of ratings allow for adjustments to personalized recommendation strategies and optimization of user experience. The number of ratings in grouped data is determined by defining criteria for user grouping, such as geographic location, age, gender, and purchase history. Based on these criteria, users are divided into corresponding groups, and user rating data is obtained. For each group, the number of ratings by users in that group is calculated. Analyzing the number of ratings in each group helps understand the level of user engagement and participation in different groups. A higher number of ratings indicates more frequent and active rating behavior in that group, while a lower number of ratings indicates less or less active rating behavior. Based on these differences in rating numbers, personalized recommendation strategies can be adjusted and user experience optimized.

[0120] `max` represents a constant; a constant refers to a fixed value or parameter used in a reference standard. Constants can be obtained by studying relevant literature, reference books, or communicating with experts, as experts and practitioners may have already identified certain constants for a specific field or problem. Alternatively, some constants may be defined and given in standards, industry guidelines, or related documents. These documents typically provide fixed values ​​or parameters used in a specific field or industry.

[0121] S v Non-first-time rating vectors for grouped data refer to the rating vectors formed for users who have already rated their data within each group after user grouping. In user behavior data analysis and personalized recommendations, users are typically grouped to better understand their preferences and behavioral characteristics. When users have already rated their data and belong to a specific group, these rating records can be grouped into a vector to represent the rating preferences of users within that group. Non-first-time rating vectors for grouped data can contain multiple dimensions, each representing a rating metric or feature. By analyzing these vectors, differences in rating preferences among users in different groups can be revealed, leading to a better understanding of users' personalized needs. The acquisition of non-first-time rating vectors for grouped data involves... The criteria for grouping users are defined, such as geographic location, age, gender, and purchase history. Based on the selected grouping criteria, users are divided into corresponding groups, and user rating data is obtained. This may involve accessing a rating database or recording system and retrieving each user's rating record in each group. For users who have already rated in each group, their rating records are combined into a rating vector. Each rating vector represents a user's rating preference in that group. The rating vector can be multi-dimensional, with each dimension representing a rating indicator or feature. The group data is analyzed to identify the differences in rating preferences among users in different groups. The similarities and differences between the rating vectors of different groups are compared to identify the common rating preferences and personalized needs of users in each group.

[0122] x iThe weighting of the number of ratings in grouped data refers to the weight assigned to the number of ratings when analyzing and calculating grouped data. The weighting of the number of ratings in grouped data can be determined by domain experts: ask domain experts or relevant stakeholders to provide subjective opinions and suggestions on the importance of the number of ratings in the analysis, and determine the weighting based on experience and professional knowledge. Consider the degree of influence of the number of ratings in grouped data on user behavior and preferences, or explore the correlation or influence between the number of ratings and other indicators (such as user satisfaction, purchasing behavior, etc.) by analyzing existing datasets. Through statistical analysis or machine learning techniques, the weighting of the number of ratings in model training or prediction can be obtained.

[0123] S32223. Input the test machine into the improved BP neural network model after training and test it. Use cross-validation to train and test the improved BP neural network model multiple times, and calculate the average error to estimate the accuracy of the result.

[0124] Specifically, after the calculation is completed, the time series model is fitted. The basic model is fitted using an appropriate method, and the model parameters are determined. The least squares method or the maximum likelihood method is used to estimate the model parameters. Then, the model is tested. The fitted model is tested, and the model's goodness of fit and prediction accuracy are evaluated. Indicators such as mean absolute error or root mean square error are used to evaluate the model's prediction accuracy.

[0125] S32223. Optimize the improved BP neural network model based on the estimation results;

[0126] S323. Calculate the objective preference value based on the grouped objective data, and assign weights to the objective preference value;

[0127] S324. Calculate the user's objective preference for the product based on the weighted time-series predicted value and objective preference value;

[0128] S33. Calculate the user's subjective preference using the de-identified subjective data;

[0129] Specifically, factor analysis algorithms are used to mathematically model multiple data points related to user product purchases to detect common structural patterns in the data, thereby reducing data dimensionality and improving interpretability. The basic idea is to infer underlying common factors from observable variables; changes in the values ​​of these common factors affect their values ​​in the observable variables, thus determining subjective preferences.

[0130] The process of calculating users' subjective preferences using anonymized subjective data includes the following steps:

[0131] S331. Conduct data analysis on subjective data;

[0132] Specifically, data analysis of subjective data includes the following steps: analyzing access paths based on subjective data to obtain users' interests, browsing habits, and behavioral patterns;

[0133] Specifically, individual data analysis involves: designing questionnaires: designing appropriate questionnaire content and questions based on the information needed, such as product features, prices, brands, and services; determining the sample: identifying the individual sample to be surveyed, such as consumers, users, and customers; distributing questionnaires: distributing the designed questionnaires to the sample individuals, either online or offline; collecting data: collecting individual responses to different questions, and organizing and storing the data; data analysis: performing statistical analysis on the collected data, such as calculating the mean, standard deviation, and frequency distribution; and interpreting and applying the results: interpreting the degree of individual preference for different questions based on the data analysis results, and applying the results to improve products or services.

[0134] The following points should be noted when using the questionnaire survey method: The questionnaire design should be reasonable: Questions should be set appropriately based on the actual situation, avoiding too many or too few questions. The sample selection should be appropriate: The sample selection should be representative and able to reflect the overall situation of the group. Data collection should be timely: Data collection should be timely to avoid omissions or invalid data. Data analysis should be accurate: Data analysis should be accurate to avoid biased results due to improper data processing. The questionnaire survey method can help businesses and organizations understand individuals' preferences and needs regarding products and services, thereby better meeting consumer needs and improving business performance.

[0135] After data collection, individual data analysis using behavioral analysis methods includes the following steps:

[0136] Define behavioral metrics: Based on the information needed, determine appropriate behavioral metrics, such as purchase frequency, purchase amount, and browsing duration. Collect behavioral data: Collect data on individual behaviors, such as purchase history and browsing history. Data cleaning: Clean the data, removing invalid data, such as duplicates and outliers. Data analysis: Analyze the collected data, such as calculating the mean, standard deviation, and frequency distribution. Interpret and apply the results: Based on the data analysis results, interpret the degree of individual preference for a certain product or service, and apply the results to improve the product or service.

[0137] The following points should be noted when using behavioral analysis: Behavioral indicators must be accurate: They must accurately reflect an individual's preferences and should not be overly simplistic or one-sided. Data collection must be complete: Data collection must be complete, avoiding missing or incomplete data.

[0138] Based on subjective data, we can analyze purchase paths to obtain users' purchase preferences, shopping habits, and decision-making processes.

[0139] Specifically, access analytics understands users' interests and preferences for different pages or functions by analyzing their behavior when visiting websites or applications. It analyzes users' attention to different pages by statistically analyzing data such as the time users spend on different pages and the number of clicks. At the same time, by analyzing users' traffic data when visiting websites or applications, it can understand the characteristics and preferences of users in different regions by analyzing users' geographic and time information.

[0140] We use subjective data to analyze user behavior and understand their needs and preferences.

[0141] Specifically, by analyzing users' conversion behavior on websites or applications, we can understand users' purchase intentions and behavioral patterns, and by analyzing users' purchasing behavior on shopping websites, we can understand users' shopping habits and preferences.

[0142] The following methods can be used for purchase analysis:

[0143] Purchase frequency: Analyzing the frequency of user purchases, such as the number of purchases per week, month, or quarter, helps understand user shopping habits and preferences. Analyzing the amount of money spent, such as average purchase amount and highest purchase amount, helps understand user shopping needs and values. Analyzing the categories of goods purchased, such as food, clothing, and home goods, helps understand user shopping preferences and needs. Analyzing the time of purchase, such as weekdays, weekends, or holidays, helps understand user shopping habits and preferences. Analyzing the channels through which users purchase, such as websites, apps, or WeChat mini-programs, helps understand user shopping preferences and habits. Analyzing user reviews and ratings of purchased products helps understand user preferences and needs for those products. Analyzing user behavior of adding and deleting items in the shopping cart helps understand user shopping decision-making processes and preferences. By analyzing user purchasing behavior on shopping websites, we can better understand user shopping needs and preferences, thereby providing products and services that better meet user needs, enhancing user satisfaction and loyalty, and promoting business growth and development.

[0144] Specifically, by analyzing user behavior paths on websites or applications, we can understand which pages users enter from, which pages they pass through, and which page they ultimately leave from. By analyzing user behavior paths, we can discover different user behavior patterns and interests, and increase the investment in links to pages that are frequently accessed and opened by URL.

[0145] S332. Calculate the subjective preference value based on the pre-analyzed subjective data, and calculate the subjective preference degree by weighting the subjective preference value.

[0146] Specifically, time series analysis models can also be constructed using the following models: Autoregressive moving average model: The autoregressive moving average model is a time series-based forecasting model that decomposes the time series into trend, seasonal, and random components and uses these components to build a forecasting model. The advantage of the autoregressive moving average model is that it can adapt to different time series patterns, but the disadvantage is that it requires experience in time series analysis and forecasting.

[0147] Seasonal Autoregressive Integrated Moving Average Model: The seasonal autoregressive integrated moving average model is an extension of the seasonal autoregressive integrated moving average model. It is better suited to time series data with seasonality and can handle non-stationary time series data.

[0148] Long Short-Term Memory Network: Long Short-Term Memory Network is an algorithm based on recurrent neural networks. It can model and predict time series data, and can handle long series data. It is suitable for prediction of multiple time steps.

[0149] Neural network autoregressive model: The neural network autoregressive model is a time series forecasting model based on neural networks. It uses data from several previous time points to predict future values ​​and can be used to predict multiple variables.

[0150] Siamese Networks: Siamese networks are a similarity measurement algorithm based on neural networks. They can be used for similarity calculation and anomaly detection of time series data. However, the construction of time series analysis models needs to be determined based on the characteristics of the data and the specific application scenario. At the same time, attention should be paid to data preprocessing and feature engineering to improve the prediction accuracy and robustness of the time series model.

[0151] S4. Combine the user's subjective and objective preferences based on the preset reliability, and calculate the user's overall preference for the product;

[0152] The process of combining users' subjective and objective preferences based on a preset reliability level and calculating users' overall preference for the product includes the following steps:

[0153] S41. Assign weights to subjective and objective preferences respectively;

[0154] S42. Calculate the overall preference score based on the weights of subjective preference and objective preference.

[0155] The formula for calculating the comprehensive preference score, based on the weighting of subjective and objective preferences, is as follows:

[0156] y = ax + zb

[0157] Where y represents the overall preference level;

[0158] 'a' represents the weight of subjective preference.

[0159] x represents the degree of subjective preference;

[0160] z represents the objective preference weight;

[0161] b represents the degree of objective preference.

[0162] Specifically, the following methods can be used to assign weights to subjective and objective preferences: equal weighting method: set the weight values ​​of all variables or indicators to be equal, that is, each variable or indicator has the same impact on the final result. This method is suitable for situations where the importance of each variable or indicator is equal.

[0163] Expert evaluation method: This method assigns appropriate weight values ​​to different variables or indicators based on the experts' experience and judgment. It is suitable for situations where there are significant differences between variables or indicators and multiple factors need to be considered.

[0164] Analytic Hierarchy Process (AHP): AHP is a method based on a judgment matrix. It compares and evaluates the relative importance of different variables or indicators to arrive at the final weight values. This method is suitable for situations where there is a complex hierarchical structure among multiple variables or indicators.

[0165] Principal Component Analysis (PCA): PCA is a method that reduces multiple variables or indicators into a few comprehensive indicators by reducing dimensionality. Through PCA, the weight values ​​of each comprehensive indicator can be obtained, thus yielding the weight values ​​of the final result.

[0166] Entropy weighting method: The entropy weighting method is a method that quantifies the importance of different variables or indicators by using information entropy. By calculating the information entropy value of different variables or indicators, the weight value of each variable or indicator can be obtained. The choice of weighting algorithm needs to consider the specific application scenario and data characteristics. At the same time, it is necessary to ensure the rationality and accuracy of the weight values ​​when assigning weights.

[0167] S5. Analyze users' purchasing preferences for products based on comprehensive preference analysis, and recommend corresponding products to users based on their comprehensive preferences.

[0168] The process of analyzing user purchase preferences based on comprehensive preference data and recommending corresponding products to users based on these preferences includes the following steps:

[0169] S51. Recommend similar items based on the user's overall product preference for items with a high overall preference level.

[0170] S52. Based on the user's overall product preference, find similar items to increase promotion efforts, and dynamically monitor the prices of similar items based on the user's overall product preference.

[0171] According to another embodiment of the present invention, such as Figure 2 As shown, a user product preference analysis system based on data security is provided. The system includes:

[0172] The data acquisition module is used to acquire users' browsing and purchase information and analyze it to obtain user behavior parameters.

[0173] The data cleaning module is used to clean the obtained user behavior parameters and classify the browsing and purchase data based on the product type to obtain categorized behavior data.

[0174] The behavioral data analysis module is used to calculate users' subjective and objective preferences for products based on categorized behavioral data.

[0175] The comprehensive preference acquisition module is used to combine users' subjective and objective preferences based on a preset reliability, and calculate users' comprehensive preference for the product.

[0176] The recommendation adjustment module is used to analyze users' purchasing preferences for products based on the calculated comprehensive preference score, and recommend corresponding products to users based on their comprehensive preferences.

[0177] In summary, by utilizing the above-mentioned technical solutions of this invention, the present invention categorizes the interpretation results of user product preference values ​​and performs data anonymization processing, which facilitates the protection of user privacy and data security. At the same time, it facilitates real-time monitoring of product sales driven by live-streaming e-commerce and influencer effects, enabling merchants to adjust product procurement in a timely manner based on user product preference analysis, thus greatly improving the efficiency of this data security-based user product preference analysis method.

[0178] Furthermore, this invention improves the accuracy of user product preference analysis methods by performing time-series data analysis to observe product sales and user purchasing preferences in a timely manner. At the same time, it recommends high-selling products to users in real time by allocating weights based on overall preference values ​​and individual preference values, thereby improving the analytical effect of user product preference analysis methods. Moreover, by analyzing individual preference values, it enhances the convenience for users to purchase items that meet their needs.

[0179] Furthermore, by finding and recommending similar items based on user product preference values, the convenience of product selection for users is improved. At the same time, recommending items that users need based on their product preference values ​​improves the sales efficiency of merchants' products. This greatly enhances the efficiency of user product preference analysis methods based on data security, and provides real-time dynamic price monitoring of products that users are interested in, further improving the convenience of user experience.

[0180] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A user product preference analysis method based on data security, characterized in that, The method includes the following steps: S1. Obtain user browsing and purchase information, and analyze it to obtain user behavior parameters; S2. Clean the obtained user behavior parameters and classify the browsing and purchase data according to the product type to obtain classified behavior data. S3. Calculate users' subjective and objective preferences for the product based on the classification behavior data; The calculation of users' subjective and objective preferences for products based on classification behavior data includes the following steps: S31. Divide the classification behavior data into subjective data and objective data, and perform anonymization processing on the subjective data and objective data; S32. Calculate the user's objective preference based on the de-identified objective data; The calculation of user objective preference based on anonymized objective data includes the following steps: S321. Group the desensitized objective data according to the product type to obtain objective grouped data; S322. Establish a time series forecasting model based on the grouping results, analyze and forecast the grouped objective data based on the time series forecasting model to obtain the time series forecast values, and assign weights to the time series forecast values. The steps of establishing a time-series prediction model based on the grouping results, analyzing and predicting the grouped objective data using the time-series prediction model to obtain time-series prediction values, and assigning weights to the time-series prediction values ​​include the following: S3221. Obtain historical data based on classification behavior data, analyze the historical data to obtain objective data, and preprocess the objective data. S3222. Construct and train an improved BP neural network model based on the preprocessed objective data; The process of constructing and training an improved BP neural network model based on preprocessed objective data includes the following steps: S32221. Use 60% of the preprocessed objective data as the training set and 40% as the test set. S32222. Select a personalized system recommendation algorithm to construct an improved BP neural network model, and train the model using the training set. The formula for the improved BP neural network model is: ； In the formula, These are time-series predicted values ​​of objective data; Similarity between objective data; A rating vector for objective data; It is a mixed weighting factor; The number of rating values; This is the first score vector for the grouped data; S32223. Input the test machine into the improved BP neural network model after training and test it. Use cross-validation to train and test the improved BP neural network model multiple times, and calculate the average error to estimate the accuracy of the result. S32223. Iterate the improved BP neural network model based on the estimation results; S3223. Real-time prediction of collected objective data based on the improved BP neural network model; S3224. Use the prediction results of the improved BP neural network model as the time-series prediction value of objective data; S323. Calculate objective preference values ​​based on objective grouping data and assign weights to the objective preference values; S324. Calculate the user's objective preference for the product based on the time-series predicted value and the objective preference value after weighting. S33. Calculate the user's subjective preference based on the de-identified subjective data; S4. Combine the user's subjective and objective preferences based on the preset reliability, and calculate the user's overall preference for the product; S5. Analyze users' purchasing preferences for products based on comprehensive preference analysis, and recommend corresponding products to users based on their comprehensive preferences.

2. The user product preference analysis method based on data security according to claim 1, characterized in that, The process of dividing classification behavior data into subjective data and objective data, and then anonymizing the subjective and objective data, includes the following steps: The anonymized subjective and objective data were categorized by region and time, and statistical analysis was performed on the anonymized subjective and objective data to obtain statistical indicators such as the mean, standard deviation, maximum and minimum values ​​of the anonymized subjective and objective data.

3. The user product preference analysis method based on data security according to claim 1, characterized in that, The calculation of user subjective preference based on anonymized subjective data includes the following steps: S331. Conduct pre-analysis of subjective data; S332. Calculate the subjective preference value based on the pre-analyzed subjective data, and calculate the subjective preference degree by weighting the subjective preference value.

4. The user product preference analysis method based on data security according to claim 3, characterized in that, The process of combining users' subjective and objective preferences based on a preset reliability level and calculating users' overall preference for the product includes the following steps: S41. Assign weights to subjective and objective preferences respectively; S42. The overall preference score is calculated based on the weights of subjective preference and objective preference.

5. The user product preference analysis method based on data security according to claim 4, characterized in that, The process of analyzing user purchase preferences based on comprehensive preference and recommending corresponding products to users based on their comprehensive preferences includes the following steps: S51. Recommend similar items based on the user's overall product preference for items with a high overall user preference. S52. Find similar items based on the user's overall product preference, increase the frequency of similar item push, and perform real-time price analysis on similar items based on the user's overall product preference, and push real-time price changes to the user.

6. A user product preference analysis system based on data security, used to implement the steps of the user product preference analysis method based on data security as described in any one of claims 1-5, characterized in that, The system includes: The data acquisition module is used to acquire users' browsing and purchase information and analyze it to obtain user behavior parameters. The data cleaning module is used to clean the obtained user behavior parameters and classify the browsing and purchase data based on the product type to obtain categorized behavior data. The behavioral data analysis module is used to calculate users' subjective and objective preferences for products based on categorized behavioral data. The comprehensive preference acquisition module is used to combine users' subjective and objective preferences based on a preset reliability, and calculate users' comprehensive preference for the product. The recommendation adjustment module is used to analyze users' purchasing preferences for products based on their overall preference score, and recommend corresponding products to users based on their overall preferences.

Images