E-commerce marketing system based on big data analysis

Abstract

The invention relates to the technical field of e-commerce, in particular to an e-commerce marketing system based on big data analysis, which comprises a user data integration module, a real-time market monitoring module, an advertisement resource optimization module and a purchase behavior prediction module. According to the method, fine insight of user behaviors is realized by analyzing big data from social media, mobile applications and retail platforms, the practicability of the data is improved, and the user portraits are accurately constructed through feature normalization and irrelevant information removal, so that commodity recommendation is more personalized and timely through fine user analysis, and the user experience is improved. The participation degree and purchase behaviors of users are effectively improved, the real-time monitoring function of the system enables commodity matching to be more sensitive, content is adjusted in time to match user requirements and market changes, dynamic optimization of advertisement resources is adjusted according to market feedback, the efficiency of the advertisement resources is maximized, return benefits are improved, and by predicting and analyzing the purchase trend of the users, the user experience is improved. And a scientific basis is provided for commodity inventory and marketing promotion.

Description

Technical Field

[0001] The present invention relates to the field of e-commerce technology, and in particular, to an e-commerce marketing system based on big data analysis. Background Art

[0002] The field of e-commerce technology involves using Internet technology to buy and sell goods or services, including consumers directly purchasing products through online stores, and enterprises conducting transactions electronically. E-commerce technology has not only changed the traditional business model but also introduced new marketing strategies and tools, such as search engine optimization (SEO), online advertising, social media marketing, and mobile commerce. The key technologies in this field also include secure payment systems, electronic shopping cart technology, order management systems, and customer relationship management (CRM) systems.

[0003] Among them, the e-commerce marketing system based on big data analysis involves using big data technology to enhance the marketing effect of e-commerce platforms. By analyzing a large amount of user data (such as purchase history, search habits, and user behavior), potential market trends and consumer preferences are identified. Its main purpose is to improve sales efficiency through precision marketing. For example, by product recommendations and customized marketing activities to increase user engagement and purchase conversion rates.

[0004] The existing technologies are insufficient in instantly capturing and processing the latest market data, resulting in untimely response to market demands, reducing the user experience and purchase willingness. The static allocation of advertising resources fails to adapt to the rapid changes in the market and user behavior, causing unreasonable resource allocation and affecting the effective reach and conversion effect of advertisements. The prediction of purchase behavior relying on outdated data in the existing technologies limits the accuracy of decision-making, making the promotion of goods lack sufficient data support, thus affecting the overall sales effect of e-commerce. Summary of the Invention

[0005] The purpose of the present invention is to solve the drawbacks existing in the prior art, and to propose an e-commerce marketing system based on big data analysis.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions: An e-commerce marketing system based on big data analysis, the system includes:

[0007] The user data integration module collects user browsing records, shopping records, and interaction behaviors based on social media, mobile application, and retail platform data, analyzes the correlation between behavior characteristics, performs feature normalization processing, eliminates redundant information, and associates the behavior data with user identifiers to obtain a user behavior feature set;

[0008] The real-time market monitoring module continuously monitors the user access channels and purchase behaviors based on the user behavior feature set, detects changes in user behavior frequencies and preferences, calculates the change amounts of channel data within multiple time periods, filters out product information that conforms to the user behavior trends, and conducts a classified analysis of the change trends to obtain market behavior dynamic indicators;

[0009] The advertising resource optimization module analyzes the display priorities of the product matching sequences based on the market behavior dynamic indicators, adjusts the allocation weights of the advertising resources according to the priorities, and counts the number of ad displays and calculates the resource usage efficiency, comparing the click frequencies and conversion values of different advertising channels to obtain advertising placement decision information;

[0010] The purchase behavior prediction module collects the placement ratios and channel conversion data based on the advertising placement decision information, generates a time series in combination with the user's historical purchase behaviors, calculates the user purchase behavior trends, divides the prediction intervals, and predicts the user's future purchase behaviors based on the intervals to obtain the purchase potential evaluation results.

[0011] The improvement of the present invention is that the analysis steps of the correlation between the behavior features are specifically as follows:

[0012] Based on the data of social media, mobile applications, and retail platforms, collect the user's browsing records, shopping records, and interaction behavior data, perform data cleaning and preliminary formatting to obtain a preliminary behavior data set;

[0013] Extract features from the preliminary behavior data set, identify the key behavior features of the browsing path depth, shopping conversion rate, and feedback interaction times to obtain a key feature data set;

[0014] Based on the key feature data set, use the Pearson correlation coefficient to analyze the correlation between the behavior features, using the formula:

[0015]

[0016] Obtain the analysis results of the behavior feature correlation, where r xy represents the correlation coefficient between the behavior features x and y, n is the sample size, ∑xy is the sum of the products of the features x and y, ∑x is the total sum of the feature x, ∑y is the total sum of the feature y, ∑x 2 is the total sum of the squares of the feature x, and ∑y 2 is the total sum of the squares of the feature y.

[0017] The improvement of the present invention is that the acquisition steps of the user behavior feature set are specifically as follows:

[0018] Based on the correlation between the behavior features, merge the user's basic identification information and identify the behavior data of different user groups, using the formula:

[0019]

[0020] Obtain the user behavior pattern recognition result, where v i represents the behavior feature vector of user i, and f ik is the score of user i on feature k, and α k represents the weight coefficient of feature k, and K is the total number of features;

[0021] Based on the user behavior pattern recognition result, delete redundant or low-correlated features, optimize the data access efficiency, and obtain the user behavior feature set.

[0022] The improvement of the present invention is that the calculation steps of the change amount of the channel data within multiple time periods are specifically as follows:

[0023] Based on the user behavior feature set, continuously monitor the user channel access data and purchase behavior, detect the changes in user behavior frequency and preference, and cluster according to the user identifier to obtain the user access time series data;

[0024] Based on the user access time series data, calculate the change amount of the time series data according to the time interval between the front and back data points, and use the formula:

[0025] Δf i,t = w t ·f i,t - f i,t-1 ;

[0026] Evaluate the change trend and abnormal points to obtain the change amount of the channel data within multiple time periods, where Δf i,t represents the change amount of the channel access frequency of user i at time t, f i,t represents the access frequency of user i at time t, f i,t-1 represents the access frequency of user i at the previous time point t - 1, and w t is the weight coefficient based on the time interval.

[0027] The improvement of the present invention is that the obtaining steps of the market behavior dynamic index are specifically as follows:

[0028] Based on the change amount of the channel data, analyze the user behavior data and commodity purchase data, determine the relevance between the commodity and the user purchase frequency and preference, and screen the commodity information that conforms to the user behavior trend to obtain the behavior-conforming commodity information;

[0029] Based on the behavior-conforming commodity information, classify the commodity and the user behavior trend, identify the market behavior patterns of different categories, and obtain the market behavior dynamic index.

[0030] The improvements of the present invention are as follows. The specific steps for statistically calculating the advertisement display times and resource utilization efficiency are as follows:

[0031] Based on the dynamic market behavior indicators, adjust the advertisement resource allocation, convert the product display priority into advertisement resource weights, and determine the allocation weights of each advertisement by analyzing the product matching sequence to obtain the adjusted result of the advertisement allocation weights.

[0032] Based on the adjusted result of the advertisement allocation weights, count the display times of each advertisement, including the total times and the display status by time period, to obtain the advertisement display statistical data.

[0033] Based on the advertisement display statistical data, use the formula:

[0034]

[0035] Calculate the utilization efficiency EG of the advertisement resources, where CY represents the advertisement click times, TG represents the advertisement display times, VG is the user participation coefficient, and BG is the advertisement cost, including the creation and placement costs.

[0036] The improvements of the present invention are as follows. The specific steps for obtaining the advertisement placement decision-making information are as follows:

[0037] Collect the total number of user clicks and conversion transaction records of the advertisement channels, and conduct comparative analysis on the collected data to evaluate the correlation between the click frequencies and conversion values of different advertisement channels to obtain the advertisement effect analysis result.

[0038] Based on the advertisement effect analysis result, formulate an advertisement placement plan, adjust the advertisement budgets, content designs, and placement times of different channels to optimize the utilization rate of advertisement resources and obtain the advertisement placement decision-making information.

[0039] The improvements of the present invention are as follows. The specific steps for obtaining the purchase potential evaluation result are as follows:

[0040] Based on the advertisement placement decision-making information, collect the placement ratios and conversion rates of multiple advertisement channels, and synchronously integrate the purchase behavior data of users to obtain the user behavior and advertisement placement data set.

[0041] Based on the user behavior and advertisement placement data set, calculate the user purchase behavior trend using the formula:

[0042]

[0043] Obtain the result of the trend analysis, where TC t represents the predicted purchase trend at time t, XC t-i represents the purchase data at time t - i, D ci is the weight coefficient, and n Cis the total number of data points;

[0044] Based on the results of the trend analysis, a prediction interval is divided, and the future purchase behavior of the user is predicted according to the interval to obtain the purchase potential evaluation result.

[0045] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0046] In the present invention, by analyzing the big data from social media, mobile applications and retail platforms, a fine insight into user behavior is achieved, the usability of data is improved, and a user portrait is accurately constructed by feature normalization and elimination of irrelevant information. This detailed user analysis makes product recommendations more personalized and timely, effectively improving user engagement and purchase behavior. The real-time monitoring function of the system makes product matching more sensitive, adjusts the content in a timely manner to match user needs and market changes, and the dynamic optimization of advertising resources is adjusted according to market feedback to maximize the efficiency of advertising resources, improve the return on investment, and provide a scientific data basis for product inventory and marketing promotion by predicting and analyzing the purchase trends of users. Brief Description of the Drawings

[0047] Figure 1 is the system flow chart of the present invention;

[0048] Figure 2 is the analysis flow chart of the correlation between behavior characteristics in the present invention;

[0049] Figure 3 is the acquisition flow chart of the user behavior feature set in the present invention;

[0050] Figure 4 is the calculation flow chart of the change amount of channel data in multiple time periods in the present invention;

[0051] Figure 5 is the acquisition flow chart of the dynamic index of market behavior in the present invention;

[0052] Figure 6 is the statistical calculation flow chart of the number of ad displays and resource utilization efficiency in the present invention;

[0053] Figure 7 is the acquisition flow chart of the advertising placement decision information in the present invention;

[0054] Figure 8 is the acquisition flow chart of the purchase potential evaluation result in the present invention. Detailed Description of the Invention

[0055] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0056] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the accompanying drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more, unless otherwise specifically defined.

[0057] Embodiment

[0058] Please refer to Figure 1 , the present invention provides a technical solution: An e-commerce marketing system based on big data analysis includes:

[0059] The user data integration module collects user browsing records, shopping records and interaction behaviors based on social media, mobile application and retail platform data, analyzes the correlation between behavior characteristics, performs feature normalization processing, eliminates redundant information, and associates behavior data with user identifiers to obtain a user behavior feature set;

[0060] The real-time market monitoring module continuously monitors user access channels and purchase behaviors based on the user behavior feature set, detects changes in user behavior frequency and preferences, calculates the change amount of channel data in multiple time periods, filters out product information that conforms to the user behavior trend, and classifies and analyzes the change trend to obtain market behavior dynamic indicators;

[0061] The advertising resource optimization module analyzes the display priority of the product matching sequence based on the market behavior dynamic indicators, adjusts the allocation weight of advertising resources according to the priority, and counts the number of ad displays and calculates the resource usage efficiency, compares the click frequency and conversion value of different advertising channels to obtain advertising placement decision information;

[0062] The purchase behavior prediction module collects the placement ratio and channel conversion data based on the advertising placement decision information, generates a time series in combination with the user's historical purchase behavior, calculates the user's purchase behavior trend, divides the prediction interval, and predicts the user's future purchase behavior based on the interval to obtain the purchase potential evaluation result.

[0063] The user behavior feature set specifically includes browsing type, purchase frequency, and interaction category. The market behavior dynamic indicators include product popularity, price changes, and consumer interest. The advertising placement decision-making information includes priority classification, resource allocation ratio, and placement effect evaluation results. The purchase potential evaluation results include potential consumption value, loyalty indicators, and pre-purchase intention.

[0064] Please refer to Figure 2 , and the analysis steps for the correlation between behavior features are specifically as follows:

[0065] Based on data from social media, mobile applications, and retail platforms, collect users' browsing records, shopping records, and interaction behavior data, perform data cleaning and preliminary formatting to obtain a preliminary behavior dataset;

[0066] Call the access interfaces of each platform one by one to obtain users' browsing records, shopping records, and interaction behavior data. Group the data preliminarily according to the source platform and establish timestamp markers. Subsequently, perform cleaning operations, including removing duplicate records, incomplete data, and data with format errors, and remove invalid records through field logic judgment. For the timestamp field, perform unified standardization processing to convert the time formats of different platforms into a consistent format. Immediately store all the obtained data according to field normalization, including field contents such as user identifiers, behavior types, behavior times, behavior durations, and identifiers of the products or services involved. Finally, store all the cleaned and formatted data for subsequent analysis to ultimately obtain a preliminary behavior dataset.

[0067] Extract features from the preliminary behavior dataset to identify key behavior features such as browsing path depth, shopping conversion rate, and feedback interaction times to obtain a key feature dataset;

[0068] Through the behavior feature extraction method, calculate each key behavior feature in sequence. First, count the number of pages continuously clicked by the user under a product category to calculate the browsing path depth, and establish an association between the result and the user identifier. Subsequently, calculate the shopping conversion rate by performing a ratio operation on the total number of shopping behavior records and the total number of browsing behavior records, and use the calculated result as the shopping conversion rate data for each user. At the same time, count the number of interaction behaviors such as comments, likes, and collections submitted by each user and record it as the feedback interaction times. During the extraction process, all calculated feature data are mapped and stored according to the user identifier. After the feature calculation is completed, associate each feature with the user identifier one by one to finally generate a key feature dataset.

[0069] Based on the key feature dataset, use the Pearson correlation coefficient to analyze the correlation between behavior features, using the formula:

[0070]

[0071] Obtain the result of behavioral feature correlation analysis, where r xy represents the correlation coefficient between behavioral features x and y, n is the sample size, i.e., the number of monitoring points in the dataset, ∑xy is the sum of the products of features x and y, ∑x is the sum of feature x, ∑y is the sum of feature y, ∑x 2 is the sum of the squares of feature x, and ∑y 2 is the sum of the squares of feature y;

[0072] The sample size in the key feature dataset is n = 2, the two sample values of the path depth are x1 = 5 and x2 = 7, and the two sample values of the shopping conversion rate are y1 = 0.3 and y2 = 0.5. Parameter calculation:

[0073] ∑x = x1 + x2 = 5 + 7 = 12;

[0074] ∑y = y1 + y2 = 0.3 + 0.5 = 0.8;

[0075] ∑xy = x1·y1 + x2·y2 = 5·0.3 + 7·0.5 = 1.5 + 3.5 = 5.0;

[0076] ∑x 2 = x 2 1 + x 2 2 = 5 2 + 7 2 = 25 + 49 = 74;

[0077] ∑y 2 = y 2 1 + y 2 2 = 0.3 2 + 0.5 2 = 0.09 + 0.25 = 0.34;

[0078] Substitute into the formula:

[0079]

[0080] Calculate the numerator:

[0081] 25.0 - 12×0.8 = 10 - 9.6 = 0.4;

[0082] Calculate the first part of the denominator:

[0083] 2×74 - 12 2 = 148 - 144 = 4;

[0084] Calculate the second part of the denominator:

[0085] 2×0.34 - 0.8 2 = 0.68 - 0.64 = 0.04;

[0086] Calculate the denominator:

[0087]

[0088] Calculate the correlation coefficient:

[0089]

[0090] Final result:

[0091] r xy = 1.0;

[0092] This result indicates that there is a perfect positive correlation between the browsing path depth and the shopping conversion rate, which means that in the sample, as the browsing path depth increases, the shopping conversion rate also increases proportionally. This result directly reflects the strong correlation between behavioral characteristics.

[0093] Please refer to Figure 3 , the specific steps for obtaining the user behavior characteristic set are as follows:

[0094] Based on the correlation between behavioral characteristics, merge the basic identification information of users, and identify the behavioral data of different user groups. Use the formula:

[0095]

[0096] to obtain the user behavior pattern recognition result. Among them, v i represents the behavioral characteristic vector of user i, f ik is the score of user i on feature k, which reflects the performance or activity level of the user on this specific behavioral characteristic. α k represents the weight coefficient of feature k, which is used to adjust the contribution degree of each feature to the user behavioral characteristic vector. K is the total number of features;

[0097] Extract the basic identification information dataset of users, remove duplicates and standardize it according to the user unique identifier to ensure that the identification information of all users has a unified format and is unique. Subsequently, map and match the processed identification information with the behavioral characteristic correlation dataset, and merge the records of the two datasets one by one through the user unique identifier. During the processing, eliminate duplicate merges and unmatched records, and finally output the complete merged dataset containing the user's basic identification information and behavioral characteristics. The total number of features K = 3, and the features include the browsing path depth, the shopping conversion rate, and the number of feedback interactions. f i1 = 5, f i2 = 0.8, f i3User i's scores on three features are 5 for browsing path depth, 0.8 for shopping conversion rate, and 10 for feedback interaction times, with weight coefficients of α1 = 0.4, α2 = 0.3, and α3 = 0.3, respectively, which are set based on the importance of the features and their contribution to the overall behavior pattern.

[0098] Substitute the parameters into the formula:

[0099] v i = 0.4·5 + 0.3·0.8 + 0.3·10;

[0100] v i = 2.0 + 0.24 + 3.0;

[0101] v i = 5.24;

[0102] This result indicates that the value of the behavior feature vector of user i is 5.24, comprehensively reflecting the user's performance in browsing path depth (accounting for 40% weight), shopping conversion rate (accounting for 30% weight), and feedback interaction times (accounting for 30% weight), providing a quantitative basis for subsequent user clustering and behavior pattern analysis.

[0103] Based on the results of user behavior pattern recognition, redundant or low-correlation features are removed to optimize data access efficiency, resulting in a user behavior feature set.

[0104] Integrate the user behavior features with their behavior pattern results. By screening and analyzing the contribution of each feature in the user feature data to the behavior pattern, calculate the correlation between each feature and the overall behavior pattern using the correlation coefficient. Mark the features with a correlation coefficient less than the preset threshold as redundant features and remove them. Then perform data compression processing on the remaining feature set, using a step-by-step dimensionality reduction method to remove redundant dimensions or duplicate features in the data, thereby reducing data redundancy. During the dimensionality reduction process, reorder the importance of the features and adjust the storage format. Then perform storage structure adjustment on the optimized data set by changing the field position order and index structure to optimize data query and access speed. Finally, verify the adjusted access efficiency by comparing the data access response times before and after optimization to obtain the user behavior feature set.

[0105] Please refer to Figure 4 , the calculation steps for the change amount of channel data within multiple time periods are specifically as follows:

[0106] Based on the user behavior feature set, continuously monitor the user's channel access data and purchase behavior, detect changes in user behavior frequency and preferences, and cluster by user identification to obtain user access time series data.

[0107] Extract the channel access data and purchase behavior records in the user behavior feature set from the monitoring device or platform interface, classify the records according to the user identification, ensure that each piece of data is bound to the corresponding user, then extract the timestamp, access frequency, and purchase behavior attributes from the classified data, organize the timestamp and the corresponding access frequency and purchase behavior and establish a multi-dimensional array for subsequent use, complete the data values of the incomplete records in the data by interpolation, replace the missing values with the median values within a reasonable range, and at the same time remove the redundant data with abnormal timestamps or purchase behavior records not conforming to the rules. Finally, store all the processed data again according to the user identification and sort it to generate a time-series user behavior data set, and obtain the user access time series data.

[0108] Based on the user access time series data, calculate the change amount of the time series data according to the time interval between the front and back data points, using the formula:

[0109] Δf i,t =w t ·f i,t -f i,t-1 ;

[0110] Evaluate the change trend and abnormal points to obtain the change amount of the channel data within multiple time periods. Among them, Δf i,t represents the change amount of the channel access frequency of user i at time t, f i,t represents the access frequency of user i at time t, which is the frequency measurement value of the user's access behavior at this time point, f i,t-1 represents the access frequency of user i at the previous time point t - 1, which is the access frequency measurement value of the user at the immediately previous time point, w t is the weight coefficient based on the time interval, which is used to adjust the influence of the change amount caused by different time intervals;

[0111] Extract the timestamp information of each user from the time series data, calculate the time interval between every two adjacent time points in chronological order, and perform a difference operation on the access frequency and purchase behavior corresponding to each time point. Calculate the change amount by subtracting the corresponding value of the previous time point from the access frequency or purchase behavior value of the current time point. Subsequently, associate the change amount with the corresponding time interval to establish preliminary statistical data on the change trend. Perform statistical processing such as mean and variance on the set of all change amounts to determine the change trend range, and at the same time screen out the abnormal points exceeding a specific standard deviation for separate marking. Finally, integrate the change trend and abnormal point information into a unified report to obtain the change amount of the channel data within multiple time periods, w t The weight coefficient of the time interval, which is obtained by calculating the reciprocal of the time interval and normalizing it, is defined as If the access frequency of user i at time point t = 2 is f i,2= 12, and is f at time point t = 1 i,1 = 8, and the time interval is t - t prev = 2 - 1 = 1, and the corresponding weight

[0112] Substitute into the formula for calculation:

[0113] Δf i,2 = w2·f i,2 - f i,1 ;

[0114] Δf i,2 = 1·12 - 8;

[0115] Δf i,2 = 1·4 = 4;

[0116] This result indicates that the change in the channel access frequency of user i at time point t = 2 relative to time point t = 1 is 4, and the value directly reflects the increment of user behavior between the two time points, providing precise data support for subsequent trend analysis.

[0117] Please refer to Figure 5 , and the specific steps for obtaining the market behavior dynamic indicators are as follows:

[0118] Based on the change amount of channel data, analyze user behavior data and commodity purchase data, determine the correlation between commodities and user purchase frequency and preferences, and screen out commodity information that conforms to the user behavior trend to obtain behavior - conforming commodity information;

[0119] First, collect user behavior data and commodity purchase data, segment the user behavior data by time series, and combine the purchase data to extract the correlation between frequently occurring commodities and purchase behaviors. Establish a behavior - commodity matrix for the user behavior data of each period, where the rows of the matrix represent different users, the columns represent different commodities, and the element values are the purchase frequencies or purchase quantities. Decompose this matrix, use statistical methods to analyze the correlation between commodities and user purchase frequencies, and screen out commodities with high correlations; Subsequently, classify the commodity data, and use fixed user behavior indicators such as purchase time interval, purchase quantity change rate, etc. to calculate the commodity indicators that match the behavior trend one by one, including change trend curve fitting and calculation of difference deviation rate, and retain the commodity data whose change trends meet specific thresholds.

[0120] Based on the behavior - conforming commodity information, classify the commodities and user behavior trends, identify market behavior patterns of different categories, and obtain market behavior dynamic indicators;

[0121] Further classify the filtered product and user behavior trend data. Set multiple categories according to the purchase frequency and the distribution of behavior trend changes. Classify the data by setting fixed classification parameters such as the purchase change rate threshold or the behavior similarity threshold. Subsequently, perform a difference analysis on the classified data. By calculating the mean variance of the purchase frequency distribution of products in different categories, user behavior indicators, etc., identify the categories with significant differences. For the identified different categories, analyze the market behavior patterns, including the performance and distribution characteristics of different products in specific user groups. At the same time, calculate the behavior pattern change rate to obtain the market behavior dynamic indicators for different categories.

[0122] Please refer to Figure 6 , and the statistical calculation steps for the number of ad displays and resource utilization efficiency are specifically as follows:

[0123] Based on the market behavior dynamic indicators, adjust the advertising resource allocation. Convert the product display priority into the advertising resource weight. By analyzing the product matching sequence, determine the allocation weight for each ad to obtain the adjusted result of the ad allocation weight.

[0124] Adjust the advertising resource allocation. Extract the product display priority data related to the market behavior indicators. Organize the data according to the product identifier and generate a sequence of product priorities. Then, adjust the weight of the advertising resource allocation according to the product priority sequence. Associate the product priority value with the advertising resource ratio. The allocation weight is calculated by weighting and calling the sales performance, user click frequency, and matching degree parameters of the product. After completing the weight calculation, generate the allocation weight corresponding to each ad and perform normalization processing to ensure the consistency of the total weight, and obtain the adjusted result of the ad allocation weight.

[0125] Based on the adjusted result of the ad allocation weight, count the number of displays of each ad, including the total number and the display status by time period, to obtain the ad display statistical data.

[0126] Count the number of displays of each ad. First, extract the original data of the number of displays from the ad delivery records, sequentially obtain the ad identifier, display time, display location, and user click records. Organize the original data into an ad display frequency table and match the data in the table according to the ad identifier in the adjusted result of the allocation weight. Subsequently, classify and summarize the number of displays as the total number of displays and the display data by time period. Calculate the average display frequency in different time periods by superimposing the display interval records of the ad to generate the ad display statistical data.

[0127] Based on the ad display statistical data, use the formula:

[0128]

[0129] Calculate the utilization efficiency EG of advertising resources, where CY represents the number of ad clicks, TG represents the number of ad impressions, VG is the user engagement coefficient calculated based on the user's stay time and interaction behavior on the ad, and BG is the advertising cost, including creation and placement costs;

[0130] The following data was collected: CY = 100 clicks, obtained through real-time data monitoring of the advertising platform; TG = 1000 impressions, obtained by accumulating each time the ad loading is completed; VG = 0.8 user engagement coefficient; BG = 2000 yuan. Substitute the values for calculation:

[0131]

[0132] The result shows that the utilization efficiency of advertising resources is -1.92, meaning that for every investment in advertising resources, in fact, considering costs and user interactions, the return on the ad is negative, indicating that the advertising strategy needs to be adjusted or the ad content optimized to improve efficiency.

[0133] Please refer to Figure 7 , and the specific steps for obtaining advertising placement decision-making information are as follows:

[0134] Collect the total number of user clicks and conversion transaction records for each advertising channel, and conduct a comparative analysis of the collected data to evaluate the correlation between the click frequencies and conversion values of different advertising channels, obtaining the advertising effect analysis results;

[0135] Classify and summarize by dimensions such as advertising channel, time period, user category, etc., store the data in a structured database table, and mark a unique advertising channel identifier. For conversion transaction records, collect the involved user behavior data, obtain the conversion transaction data through the transaction record interface of the e-commerce platform, extract core indicators such as the number of orders and conversion amount, perform matching processing on the data, associate the user click behavior with the conversion behavior, match according to the timestamp and the user's unique identifier, and deduplicate and clean the cross-channel data, including removing duplicate records, outliers, and unavailable records. Subsequently, generate a click frequency distribution table and a conversion record summary table for the advertising channels based on the cleaned data, construct a comparison table of user click frequencies and conversion values for different advertising channels, and analyze the correlation between the two by constructing a comparison model to generate the advertising effect analysis results.

[0136] Based on the advertising effect analysis results, formulate an advertising placement plan, adjust the advertising budget, content design, and placement time for different channels, optimize the utilization rate of advertising resources, and obtain the advertising placement decision-making information;

[0137] Select advertising channels with relatively high relevance as the optimization targets, extract the list of advertising channels with poor effects and their corresponding click frequency and conversion rate data. For the part of adjusting the advertising budget, calculate the budget reallocation ratio based on the conversion rate of each channel and regenerate the advertising budget allocation table according to the channel priority. For the optimization of content design, extract common user interest points and browsing behavior records according to the user characteristics of the channel, generate the direction of optimized content through the keyword extraction algorithm, and record the content adjustment suggestions in the optimization task table. For the optimization of the delivery time, perform a clustering analysis on the time distribution of the conversion records, count the click frequency distribution of the high-conversion time periods of each advertising channel, extract the high-conversion time period data and generate a delivery time adjustment plan table, and combine the high-frequency time periods to adjust the delivery time settings for each channel to form a new advertising delivery plan and resource allocation decision information.

[0138] Please refer to Figure 8 , and the specific steps for obtaining the purchase potential evaluation result are as follows:

[0139] Based on the advertising delivery decision information, collect the delivery ratio and conversion rate of multiple advertising channels, and synchronously integrate the purchase behavior data of users to obtain the user behavior and advertising delivery data set;

[0140] Extract the delivery ratio of each channel. The delivery ratio data includes details such as time period, region, and advertising type. At the same time, obtain the purchase behavior data of users from the user behavior analysis. The data records the number of purchases, types of goods, and consumption amounts of users in different time periods. Associate and match all data according to user identification, time label, and advertising channel, clean and supplement incomplete or inconsistent data, and form a complete user behavior and advertising delivery data set by calculating the conversion rate of each channel and integrating it with the user purchase behavior data.

[0141] Based on the user behavior and advertising delivery data set, calculate the user purchase behavior trend, using the formula:

[0142]

[0143] Get the result of the trend analysis, where TC t represents the predicted purchase trend at time t, XC t-i represents the purchase data at time t-i, D ci is the weight coefficient, n C is the total number of data points;

[0144] If there is the following data, n C = 5, the total number of historical data points considered, D ci weight coefficient, which is 0.1, 0.15, 0.2, 0.25, 0.3, and the weight increases with the distance in time, reflecting that the most recent data has a greater impact on the trend, XCt-i Historical purchase data, which are 100, 120, 140, 160, and 180 respectively, are substituted into the formula for calculation:

[0145]

[0146]

[0147] TC t = 150;

[0148] This result indicates that the purchase trend of the user at time t is 150, showing that according to historical data and weight allocation, the user's purchase behavior shows an upward trend, providing a quantitative basis for subsequent dynamic analysis of market behavior and resource allocation.

[0149] Based on the results of trend analysis, a prediction interval is divided, and the user's future purchase behavior is predicted according to the interval to obtain the purchase potential evaluation result;

[0150] Divide the prediction interval and predict the user's future purchase behavior according to the interval, extract the trend value and the cycle value, divide the cycle value into different time periods, divide the interval for each trend change, and then analyze the user's purchase behavior by combining the user's historical behavior and the periodic change trend. Determine the future high-purchase potential time interval according to the trend value of different time periods. At the same time, classify different user groups according to their characteristics, and further match the user group with its corresponding potential products within each prediction interval, so as to form the result of the user's future purchase behavior based on interval prediction.

[0151] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical content of the technical solution of the present invention, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention still belong to the protection scope of the technical solution of the present invention.

Claims

1. An e-commerce marketing system based on big data analysis, characterized in that: The system comprises: The user data integration module collects user browsing history, shopping history and interactive behavior based on social media, mobile applications and retail platform data, analyzes the correlation between behavioral features, performs feature normalization, removes redundant information, and associates behavioral data with user identification to obtain a user behavior feature set; The real-time market monitoring module continuously monitors user access channels and purchasing behaviors based on the user behavior feature set, detects changes in user behavior frequency and preferences, calculates changes in channel data over multiple time periods, screens commodity information that matches user behavior trends, and classifies and analyzes change trends to obtain market behavior dynamic indicators; The advertising resource optimization module analyzes the display priority of the product matching sequence based on the market behavior dynamic indicators, adjusts the allocation weight of advertising resources according to the priority, counts the number of advertising displays and calculates the resource utilization efficiency, compares the click frequency and conversion value of different advertising channels, and obtains advertising placement decision information; The purchase behavior prediction module collects the advertising delivery ratio and channel conversion data based on the advertising delivery decision information, generates a time series based on the user's historical purchase behavior, calculates the user's purchase behavior trend, divides the prediction interval, and predicts the user's future purchase behavior based on the interval to obtain the purchase potential evaluation result.

2. The e-commerce marketing system based on big data analysis according to claim 1 is characterized in that: The analysis steps of the correlation between the behavioral features are specifically as follows: Based on social media, mobile applications and retail platform data, collect users' browsing history, shopping history and interactive behavior data, perform data cleaning and preliminary formatting, and obtain a preliminary behavioral data set; Extracting features from the preliminary behavioral data set, identifying key behavioral features of browsing path depth, shopping conversion rate, and feedback interaction times, and obtaining a key feature data set; Based on the key feature data set, the Pearson correlation coefficient is used to analyze the correlation between behavioral features, using the formula: The behavior feature correlation analysis results are obtained, where r xy represents the correlation coefficient between behavioral features x and y, n is the sample size, ∑xy is the sum of the products of features x and y, ∑x is the sum of feature x, ∑y is the sum of feature y, ∑x 2 is the sum of the squares of feature x, ∑y 2 is the sum of the squares of the feature y.

3. The e-commerce marketing system based on big data analysis according to claim 1, characterized in that: The steps of obtaining the user behavior feature set are specifically as follows: Based on the correlation between the behavioral characteristics, the basic identification information of the users is merged, and the behavioral data of different user groups are identified, using the formula: Get the user behavior pattern recognition result, where v i represents the behavioral feature vector of user i, f ik is the score of user i on feature k, α k Represents the weight coefficient of feature k, where K is the total number of features; Based on the user behavior pattern recognition result, redundant or low-relevance features are deleted, data access efficiency is optimized, and a user behavior feature set is obtained.

4. The e-commerce marketing system based on big data analysis according to claim 1, characterized in that: The specific steps for calculating the change amount of the channel data in the multiple time periods are: Based on the user behavior feature set, continuously monitor user channel access data and purchase behavior, detect user behavior frequency and preference changes, and cluster by user identifier to obtain user access time series data; Based on the user access time series data, the change amount of the time series data is calculated according to the time interval between the previous and next data points, using the formula: Δf i,t =w t ·(f i,t -f i,t-1 ); Evaluate the change trend and abnormal points to obtain the change amount of channel data in multiple time periods, where Δf i,t represents the change in the channel access frequency of user i at time t, f i,t represents the access frequency of user i at time t, f i,t-1 represents the access frequency of user i at the previous time point t-1, w t is a weighting factor based on the time interval.

5. The e-commerce marketing system based on big data analysis according to claim 1, characterized in that: The steps for obtaining the market behavior dynamic indicator are as follows: Based on the change amount of the channel data, the user behavior data and the product purchase data are analyzed to determine the correlation between the product and the user's purchase frequency and preference, and the product information that matches the user's behavior trend is screened to obtain behavior-matching product information; Based on the behavior-matching product information, the products and user behavior trends are classified, the market behavior patterns of different categories are identified, and the market behavior dynamic indicators are obtained.

6. The e-commerce marketing system based on big data analysis according to claim 1, characterized in that: The statistical calculation steps of the advertisement display times and resource utilization efficiency are specifically as follows: Based on the market behavior dynamic indicators, the advertising resource allocation is adjusted, the product display priority is converted into the advertising resource weight, and the allocation weight of each advertisement is determined by analyzing the product matching sequence to obtain the advertising allocation weight adjustment result; Based on the adjustment result of the advertisement allocation weight, the number of times each advertisement is displayed is counted, including the total number of times and the display status in different time periods, to obtain advertisement display statistics; Based on the ad impression statistics, the formula is used: Calculate the utilization efficiency EG of advertising resources, where CY represents the number of ad clicks, TG represents the number of ad impressions, VG is the user engagement coefficient, and BG is the advertising cost, including the creation and delivery costs.

7. The e-commerce marketing system based on big data analysis according to claim 1, characterized in that: The steps for obtaining the advertising placement decision information are specifically as follows: Collect the total number of user clicks and conversion records of advertising channels, and conduct comparative analysis on the collected data to evaluate the correlation between click frequency and conversion value of different advertising channels, and obtain advertising effect analysis results; Based on the advertising effect analysis results, an advertising delivery plan is formulated, the advertising budget, content design and delivery time of different channels are adjusted, the utilization rate of advertising resources is optimized, and advertising delivery decision information is obtained.

8. The e-commerce marketing system based on big data analysis according to claim 1, characterized in that: The steps for obtaining the purchase potential evaluation result are specifically as follows: Based on the advertising delivery decision information, the delivery ratios and conversion rates of multiple advertising channels are collected, and the user's purchase behavior data is simultaneously integrated to obtain a user behavior and advertising delivery data set; Based on the user behavior and advertising data set, the user purchase behavior trend is calculated using the formula: The results of trend analysis are obtained, among which TC t represents the predicted purchase trend at time t, XC t-i represents the purchase data at time ti, D ci is the weight coefficient, n C is the total number of data points; Based on the results of the trend analysis, the prediction intervals are divided, and the user's future purchasing behavior is predicted based on the intervals to obtain a purchasing potential evaluation result.

Images