Intelligent recommendation system integrating big data and AI

Abstract

The invention relates to the technical field of data processing, in particular to an intelligent recommendation system fusing big data and AI, and the system comprises a scheme obtaining module which obtains a plurality of target marketing schemes; the scheme analysis module analyzes a plurality of target marketing schemes to determine a plurality of marketing features; the data acquisition module collects a plurality of historical behavior data of a user; the data analysis module analyzes a plurality of historical behavior data to determine user characteristics; the association determination module analyzes association between the user features and the plurality of marketing features based on the initial association AI model, and determines prediction priorities of the plurality of target marketing schemes based on an association analysis result; a recommendation module constructs a user recommendation page based on the predicted priority, and determines target priorities of a plurality of target marketing schemes based on the real-time behavior data of the user; and the feedback module compares a comparison result of the prediction priority and the target priority to adjust the initial association AI model to obtain a target association AI model. According to the invention, the accuracy of personalized marketing scheme recommendation of the user is improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to an intelligent recommendation system integrating big data and AI. Background Art

[0002] Traditional marketing recommendation methods are often based on experience, simple market segmentation or general marketing strategies, lacking in-depth understanding of each consumer's individual characteristics and real-time needs. By combining big data with AI technology, companies can build intelligent recommendation systems to achieve personalized recommendations for consumers.

[0003] A patent document with Chinese patent application number CN118350891A discloses a marketing campaign recommendation method and device, the method comprising: in response to the operation of a target user entering a target page, obtaining a user portrait, user behavior information and marketing campaign information of the target user; inputting the user portrait, user behavior information and marketing campaign information of the target user into a trained prediction model to obtain the target user's achievement probability for each marketing campaign, the achievement probability being the probability that the target user achieves a preset target of the marketing campaign; and recommending to the target user a marketing campaign whose achievement probability meets preset conditions.

[0004] In the prior art, only the target user's probability of achieving the target for each marketing activity is used as the basis for recommendation, and the probability of achieving the target is defined as the probability of achieving the preset target of the marketing activity. This recommendation standard is relatively simple, resulting in inaccurate personalized recommendation results. Summary of the invention

[0005] To this end, the present invention provides an intelligent recommendation system that integrates big data and AI, which can solve the problem of inaccurate recommendations for personalized marketing plans for users.

[0006] To achieve the above objectives, the present invention provides an intelligent recommendation system integrating big data and AI, the system comprising: A program acquisition module is used to obtain several target marketing programs for any target product; A scheme analysis module, connected to the scheme acquisition module, for analyzing a number of the target marketing schemes, obtaining a number of keywords and a number of data indicators, and determining a number of marketing features based on the keywords and the data indicators; A data acquisition module is used to collect some historical behavior data of users based on a preset period; a data analysis module connected to the data acquisition module, configured to analyze the historical behavior data, obtain the user's price sensitivity based on the analysis results of the purchase records, determine the user's purchase tendency based on the evaluation records, and determine the user's characteristics based on the user's price sensitivity and the user's purchase tendency; A correlation determination module, connected to the solution analysis module and the data analysis module, for analyzing the correlation between the user characteristics and the marketing characteristics based on the initial correlation AI model, and determining the prediction priority of the target marketing solutions based on the correlation analysis results; A recommendation module, connected to the association determination module, is used to determine the display content of the marketing plan page, build a user recommendation page based on the predicted priority and the page display content, collect real-time behavior data of users when browsing the recommendation page, and determine the target priority of several target marketing plans based on the real-time user behavior data; A feedback module is connected to the recommendation module and is used to adjust the initial association AI model based on the comparison result between the predicted priority and the target priority to obtain a target association AI model.

[0007] Furthermore, the solution analysis module includes: A keyword determination unit, configured to perform word segmentation processing on any of the target marketing plans to obtain a number of actual word segments, and determine a number of keywords based on word frequency statistics of the number of actual word segments and matching results between the number of actual word segments and a preset marketing vocabulary; A data indicator determination unit, used to extract a number of actual data from any of the target marketing plans, determine the data types corresponding to the number of actual data, determine the corresponding data indicators according to the data types, and obtain a number of data indicators; The marketing feature determination unit is connected to the keyword determination unit and the data indicator determination unit, and is used to determine the marketing feature based on a plurality of the keywords and a plurality of the data indicators.

[0008] Furthermore, the data indicator determination unit includes: A type determination subunit, used to determine the data type based on a data format corresponding to any of the actual data and a connecting word connected thereto; The indicator determination subunit is connected to the type determination subunit and is used to compare the data type with a preset data type in a preset indicator calculation library, perform calculation based on the comparison result, and obtain the data indicator.

[0009] Furthermore, the data acquisition module includes: A recording unit is used to record a number of historical data corresponding to the user on the product platform; The determination unit is connected to the recording unit and is used to extract a plurality of purchase records and a plurality of evaluation records from a plurality of historical data within a preset period.

[0010] Furthermore, the data analysis module includes: a sensitivity analysis unit, configured to determine the purchase quantity of the target product and the corresponding purchase amount of the target product purchased by the user based on the purchase records, and determine the user's price sensitivity based on the purchase quantity and the corresponding purchase amount; a tendency analysis unit, for determining a number of actual satisfactions of the user to the target product based on the number of evaluation records, drawing a satisfaction change graph based on the number of actual satisfactions, analyzing the satisfaction change graph, and determining the user's purchase tendency according to the analysis result; The user feature determination unit is connected to the sensitivity analysis unit and the tendency analysis unit, and is used to determine the user feature according to the user's price sensitivity and the user's purchase tendency.

[0011] Furthermore, the tendency analysis unit comprises: A rating acquisition subunit is used to acquire a number of historical purchase ratings of the user and determine the initial satisfaction level based on the historical purchase ratings; a vocabulary analysis subunit, connected to the score acquisition subunit, for acquiring evaluation content corresponding to any of the historical purchase scores, analyzing the evaluation content based on the sentiment analysis model, identifying and determining positive and negative words in the evaluation content according to the analysis results, respectively calculating the proportion of positive and negative words in the evaluation content, obtaining the positive proportion and the negative proportion, comparing the positive proportion with the negative proportion, and calculating a correction coefficient based on the comparison result; a calibration subunit, connected to the vocabulary analysis subunit, for calibrating the initial satisfaction level based on the correction coefficient to obtain an actual satisfaction level; A curve drawing subunit, connected to the calibration subunit, for drawing the satisfaction change graph based on a number of the actual satisfaction levels; The tendency analysis subunit is connected to the curve drawing subunit and is used to analyze the curve change trend corresponding to the satisfaction change graph and determine the user's purchase tendency degree according to the trend change analysis result.

[0012] Furthermore, the tendency analysis subunit includes: A tendency analysis block is used to identify the change trend of the curve, so as to determine the user's purchase tendency according to the trend change analysis result; The degree analysis block is connected to the tendency analysis block and is used to calculate the slope of the curve in the satisfaction change graph to determine the tendency degree according to the slope calculation result.

[0013] Furthermore, the association determination module includes: A model building unit, used to build an initial correlation AI model based on a number of historical marketing feature information and a number of corresponding historical user feature information; A correlation determination unit, connected to the model building unit, for determining the correlation between the user feature and a plurality of the marketing features based on the initial correlation AI model; The priority determination unit is connected to the relevance determination unit and is used to determine the prediction priority based on the sorting results of the several relevance degrees.

[0014] Furthermore, the recommendation module includes: A content determination unit, configured to use the marketing features corresponding to the target marketing plans as primary display content and the entire content of the target marketing plans as secondary display content; a page construction unit connected to the content determination unit, configured to determine a page area of ​​a user recommendation page, determine a plurality of target area areas and an arrangement order of a plurality of target marketing plans based on the predicted priority, and display a plurality of primary display contents on the user recommendation page based on the arrangement order and the plurality of target area areas; The real-time adjustment unit is connected to the page construction unit and is used to determine the target priority based on the real-time click count of the user's primary display content and the real-time browsing time and real-time purchasing behavior of the corresponding secondary display content.

[0015] Furthermore, the feedback module includes: A comparison unit, used to compare the actual difference between the target priority and the predicted priority; A model adjustment unit is connected to the comparison unit and is used to adjust the initial association AI model based on the comparison result between the actual difference and the preset difference, or to reconstruct the association AI model to obtain a target association AI model.

[0016] Compared with the prior art, the beneficial effect of the present invention lies in that, by setting the scheme acquisition module to collect a variety of target marketing schemes, the recommendation system can screen and match from a large number of marketing schemes of different types, different discount strengths, and different activity forms when making recommendations to users in the future, thereby increasing the richness and diversity of the recommended content and meeting the diverse needs of different users for marketing activities. By setting the scheme analysis module to conduct a detailed analysis of the acquired target marketing scheme, it is possible to accurately determine a number of marketing features therein, intuitively reflect the core advantages and characteristics of the marketing scheme, and facilitate subsequent accurate matching with user characteristics. By setting the data acquisition module and the data analysis module to analyze the collected user historical behavior data, a variety of user characteristics are determined, the user's consumption behavior pattern and psychological characteristics are portrayed, and each user is personalized. Characterization and classification are performed, which provides the possibility for subsequent differentiated recommendations for different user groups, and improves the accuracy of recommendations and user satisfaction. By setting the association determination module to analyze the correlation between user characteristics and marketing characteristics based on the initial association AI model, it is possible to find out which marketing scheme characteristics are related to specific user groups. The characteristics of the body are most matched, so that the recommendation system can accurately screen out the marketing plans that are most likely to attract users according to the user's personality characteristics, improve the relevance of the recommendation, and by analyzing the correlation and determining the prediction priority of several target marketing plans, it can reasonably sort many marketing plans, further improving the user's attention to the recommended content. By setting the recommendation module, according to the degree of match between different marketing plans and users, the display position and area size of the marketing plans on the page are reasonably arranged, providing personalized visual presentation, improving the user experience and acceptance of recommended content. By determining the target priority of several target marketing plans based on the user's real-time behavior data, the recommendation strategy can be adjusted in time according to the user's current actual behavior, so that the recommendation is more in line with the user's real-time needs, further improving the accuracy and effectiveness of the recommendation. By setting the feedback module to compare the comparison results of the prediction priority and the target priority to adjust the initial association AI model, the model can be continuously optimized according to the difference between the user's actual behavior and the model prediction, so that the model can better adapt to market changes, changes in user needs and changes in marketing plan characteristics, and maintain the accuracy and effectiveness of the recommendation system.

[0017] In particular, by setting up a keyword determination unit to perform word segmentation on the target marketing plan and combining word frequency statistics and matching with a preset marketing vocabulary, the key information can be accurately extracted, which is helpful for subsequent faster and more accurate understanding of the key content of the marketing plan, laying the foundation for further analysis and matching of user needs, converting the marketing plan into a series of keywords, realizing standardized processing of marketing copy, facilitating unified analysis, comparison and processing within the system, and improving the efficiency and accuracy of the entire recommendation system in processing marketing plan information. By setting up a data indicator determination unit to determine the corresponding data indicators, various specific details in the marketing plan can be quantitatively reflected, providing reliable data support for subsequent accurate analysis of the attractiveness, cost-effectiveness and matching degree of the marketing plan with user characteristics, avoiding the limitations of relying solely on fuzzy descriptions for judgment, and by setting up a feature determination unit to comprehensively and comprehensively characterize the marketing characteristics of the marketing plan, so that when matching the marketing plan with the user characteristics, it is possible to more accurately find the plan that best matches the user's needs and preferences, thereby improving the ability and accuracy of the recommendation system to recommend suitable marketing plans to users.

[0018] In particular, by setting up a model construction unit to construct an initial correlation AI model, a solid foundation is laid for the subsequent accurate analysis of the correlation between user characteristics and marketing characteristics, so that the recommendation system can make more reasonable and accurate correlation judgments based on historical experience, which helps to improve marketing effectiveness. By setting up a correlation determination unit based on the constructed initial correlation AI model, the correlation between user characteristics and marketing characteristics can be accurately determined, which provides an important basis for the subsequent reasonable arrangement of the recommendation order of marketing plans, which helps to improve the accuracy of recommendations. By setting up a priority determination unit, the recommendation priority of each marketing plan is clarified for the recommendation system, so that when displaying marketing plans to users, they are arranged in order from high to low priority, ensuring that users first see the marketing plan that is most likely to meet their interests and needs, making the recommended content more in line with the actual needs of users, thereby improving the effectiveness and accuracy of the recommendation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A structural block diagram of an intelligent recommendation system integrating big data and AI provided in an embodiment of the present invention; Figure 2 A structural block diagram of a solution analysis module in an intelligent recommendation system integrating big data and AI provided in an embodiment of the present invention; Figure 3 A structural block diagram of a data analysis module in an intelligent recommendation system integrating big data and AI provided in an embodiment of the present invention; Figure 4 This is a structural block diagram of the association determination module in the intelligent recommendation system integrating big data and AI provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with 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.

[0021] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0022] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be understood as a limitation on the present invention.

[0023] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0024] See also Figure 1 As shown, an embodiment of the present invention provides an intelligent recommendation system integrating big data and AI, the system comprising: A plan acquisition module 10 is used to acquire several target marketing plans for any target product; A scheme analysis module 20, connected to the scheme acquisition module 10, for analyzing a number of the target marketing schemes, obtaining a number of keywords and a number of data indicators, and determining a number of marketing features based on the keywords and the data indicators; A data acquisition module 30, used to collect a number of historical behavior data of the user based on a preset period; A data analysis module 40, connected to the data acquisition module 30, for analyzing the historical behavior data, obtaining the user's price sensitivity based on the analysis results of the purchase records, and determining the user's purchase tendency based on the evaluation records, and determining the user's characteristics based on the user's price sensitivity and the user's purchase tendency; The association determination module 50 is connected to the solution analysis module 20 and the data analysis module 40, and is used to analyze the association between the user characteristics and the marketing characteristics based on the initial association AI model, and determine the prediction priority of the target marketing solutions based on the association analysis results; The recommendation module 60 is connected to the association determination module 50, and is used to determine the display content of the marketing plan page, build a user recommendation page based on the predicted priority and the page display content, and collect real-time behavior data of users when browsing the recommendation page, and determine the target priority of several target marketing plans based on the real-time user behavior data; The feedback module 70 is connected to the recommendation module 60 and is used to compare the predicted priority and the target priority to adjust the initial association AI model and obtain the target association AI model.

[0025] Specifically, the embodiment of the present invention sets the scheme acquisition module to collect a variety of target marketing schemes, so that when the recommendation system makes recommendations to users in the future, it can screen and match from a large number of marketing schemes of different types, different discount strengths, and different activity forms, thereby increasing the richness and diversity of the recommended content and meeting the diverse needs of different users for marketing activities. By setting the scheme analysis module to perform a detailed analysis on the acquired target marketing schemes, it is possible to accurately determine a number of marketing features therein, intuitively reflect the core advantages and characteristics of the marketing schemes, and facilitate subsequent precise matching with user characteristics. By setting the data acquisition module and the data analysis module to analyze the collected user historical behavior data, a variety of user characteristics are determined, the user's consumption behavior patterns and psychological characteristics are portrayed, and each user is personalized. Characterization and classification are performed, which makes it possible to make differentiated recommendations for different user groups in the future, thereby improving the accuracy of recommendations and user satisfaction. By setting the association determination module to analyze the correlation between user characteristics and marketing characteristics based on the initial association AI model, it is possible to find out which marketing scheme characteristics are related to the characteristics of a specific user group. By analyzing the correlation and determining the prediction priority of several target marketing plans, the recommendation system can reasonably sort many marketing plans, further improving the user's attention to the recommended content. By setting up the recommendation module, according to the degree of match between different marketing plans and users, the display position and area size of the marketing plans on the page are reasonably arranged, providing personalized visual presentation, improving the user experience and acceptance of recommended content. By determining the target priority of several target marketing plans based on the user's real-time behavior data, the recommendation strategy can be adjusted in time according to the user's current actual behavior, so that the recommendation is more in line with the user's real-time needs, further improving the accuracy and effectiveness of the recommendation. By setting up a feedback module to compare the comparison results of the prediction priority and the target priority to adjust the initial association AI model, the model can be continuously optimized according to the difference between the user's actual behavior and the model prediction, so that the model can better adapt to market changes, changes in user needs and changes in the characteristics of marketing plans, and maintain the accuracy and effectiveness of the recommendation system.

[0026] See also Figure 2 As shown, the solution analysis module 20 includes: A keyword determination unit 21 is used to perform word segmentation processing on any of the target marketing plans to obtain a number of actual word segments, and determine a number of keywords based on word frequency statistics of the number of actual word segments and matching results between the number of actual word segments and a preset marketing vocabulary; A data indicator determination unit 22 is used to extract a number of actual data from any of the target marketing plans, determine the data types corresponding to the number of actual data, determine the corresponding data indicators according to the data types, and obtain a number of data indicators; The marketing feature determination unit 23 is connected to the keyword determination unit 21 and the data indicator determination unit 22, and is used to determine the marketing feature based on a plurality of the keywords and a plurality of the data indicators.

[0027] Specifically, the embodiment of the present invention can accurately extract key information from the target marketing plan by setting a keyword determination unit to perform word segmentation processing on the target marketing plan and combine word frequency statistics and matching with a preset marketing vocabulary, which is helpful for subsequent faster and more accurate understanding of the key content of the marketing plan, laying a foundation for further analysis and matching of user needs, converting the marketing plan into a series of keywords, realizing standardized processing of marketing copy, facilitating unified analysis, comparison and processing within the system, and improving the efficiency and accuracy of the entire recommendation system in processing marketing plan information. By setting a data indicator determination unit to determine the corresponding data indicators, various specific details in the marketing plan are quantitatively reflected, providing reliable data support for subsequent accurate analysis of the attractiveness, cost-effectiveness and matching degree of the marketing plan with user characteristics, avoiding the limitations of relying solely on fuzzy descriptions for judgment, and comprehensively and comprehensively describing the marketing characteristics of the marketing plan by setting a feature determination unit, so that when matching the marketing plan with the user characteristics, the plan that best matches the user's needs and preferences can be found more accurately, improving the ability and accuracy of the recommendation system to recommend suitable marketing plans to users.

[0028] It can be understood that the keyword determination unit 21 described in the embodiment of the present invention can perform word segmentation processing on any of the target marketing plans by using a word segmentation tool, wherein the word segmentation tool can be jieba word segmentation, etc. For example, "100 off for purchases over 500 and a free exquisite gift" is processed as "over", "500", "off", "100", "free", and "exquisite gift".

[0029] It is understandable that the preset marketing vocabulary described in the embodiment of the present invention collects a large amount of marketing copywriting through big data, and extracts common marketing vocabulary from it, such as "discount", "gift", "full discount", etc. At the same time, the vocabulary can also be customized and expanded according to industry characteristics and business needs.

[0030] It can be understood that the keyword determination unit 21 in the embodiment of the present invention includes: A word frequency counting subunit is used to count the occurrence frequencies corresponding to the actual participles, compare the occurrence frequencies with the preset frequencies, mark the actual participles corresponding to the occurrence frequencies greater than the preset frequencies, and obtain a plurality of first marked participles; A matching subunit, used for matching the actual participles with the preset participles in the preset marketing vocabulary, marking the actual participles whose matching result similarity is greater than the preset similarity, and obtaining a plurality of second marked participles; The keyword determination subunit is connected to the word frequency statistics subunit and the matching subunit, and is used to take the words corresponding to the intersection of a number of the first marked words and a number of the second marked words as keywords to obtain a number of the keywords.

[0031] It can be understood that the preset frequency in the embodiment of the present invention is the average of several occurrence frequencies; The preset similarity is 90%.

[0032] Specifically, the data indicator determination unit includes: A type determination subunit, used to determine the data type based on a data format corresponding to any of the actual data and a connecting word connected thereto; The indicator determination subunit is connected to the type determination subunit and is used to compare the data type with a preset data type in a preset indicator calculation library, perform calculation based on the comparison result, and obtain the data indicator.

[0033] It is understandable that the type determination subunit described in the embodiment of the present invention determines the data type by analyzing the format of the actual data, such as numbers, percentages, currency symbols, etc., and connecting words such as "full", "reduction", "gift", etc. For example, if the actual data is "500" and the connecting word is "full", it can be determined that the data type is "full reduction amount"; if the actual data is "100" and the connecting word is "reduction", it can be determined that the data type is "reduction amount"; if the actual data is "2023", "12", "1", and the connecting words are "year", "month", and "day", it can be determined that the data type is "activity time"; if the connecting word contains "to", the activity time before "to" is used as the start time, and the activity time after "to" is used as the end time.

[0034] It can be understood that the preset indicator calculation library described in the embodiment of the present invention defines the mapping relationship between different data types such as full-discount amount, discount amount, activity time, etc. and corresponding indicators, and provides a specific calculation method for each data type, such as how to calculate the discount rate from the full-discount amount (the discount rate is the discount amount / full-discount amount), how to calculate the duration of the promotion activity from the activity time (activity end time - activity start time), etc. For example, for the "100 off for purchases over 500", the type determination subunit identifies "500" as the "full-discount amount" type and "100" as the "discount amount" type. The indicator determination subunit calculates the discount rate as 20% (100 / 500) based on the preset indicator calculation library.

[0035] It can be understood that the marketing feature determination unit described in the embodiment of the present invention can associate several data indicators corresponding to several of the keywords one by one, use the data indicators as the quantitative values ​​of the several keywords, compare the several quantitative values ​​with their corresponding preset values, and determine the marketing feature based on the comparison result. For example, several keywords are determined to be "full discount" and "limited time", and the data indicator determination unit recognizes that the "full discount amount" is 800 and the "discount amount" is 200. The indicator determination subunit calculates the discount rate as 25%, and the feature determination unit associates the "full discount" keyword with the discount rate of 25%. Here, the discount rate of 25% is the quantitative value of the "full discount" keyword, which is compared with the preset value. Assuming that the preset full discount attractiveness discount rate threshold is 20%, due to 2 5%>20%, it is determined that the marketing plan has a highly attractive marketing feature in terms of "full discount", for example, it can be marked as "high full discount strategy", and the data indicator determination unit determines that the activity time is from 0:00 on November 1, 2024 to 23:59 on November 3, 2024, and calculates that the activity duration is 3 days. The feature determination unit associates the "limited time" keyword with the activity duration of 3 days. 3 days is the quantitative value of the "limited time" keyword, which is compared with the preset activity duration threshold assumed to be 5 days. Since 3 days<5 days, it is determined that the marketing plan has a short-term and high-intensity marketing feature in terms of "limited time", such as being marked as "high limited time activity", where the setting of several preset values ​​is determined according to the average of the quantitative values ​​corresponding to several historical marketing activities.

[0036] Specifically, the data acquisition module 30 includes: A recording unit is used to record a number of historical data corresponding to the user on the product platform; The determination unit is connected to the recording unit and is used to extract a plurality of purchase records and a plurality of evaluation records from a plurality of historical data within a preset period.

[0037] It can be understood that the preset period in the embodiment of the present invention is one year.

[0038] See also Figure 3 As shown, the data analysis module 40 includes: A sensitivity analysis unit 41, configured to determine the purchase quantity of the target product and the corresponding purchase amount of the target product purchased by the user based on the purchase records, and determine the user's price sensitivity based on the purchase quantity and the corresponding purchase amount; a tendency analysis unit 42, configured to determine a number of actual satisfactions of the user to the target product based on the number of evaluation records, draw a satisfaction change graph based on the number of actual satisfactions, and analyze the satisfaction change graph to determine the user's purchase tendency according to the analysis result; The user feature determination unit 43 is connected to the sensitivity analysis unit 41 and the tendency analysis unit 42, and is used to determine the user feature according to the user's price sensitivity and the user's purchase tendency.

[0039] Specifically, the embodiment of the present invention sets a sensitivity analysis unit to quantitatively determine the user's price sensitivity based on the purchase quantity and the corresponding purchase amount in the purchase record, so that the system can clearly understand each user's reaction to the price change of the product, which provides an important basis for the subsequent accurate marketing recommendations and personalized services. By setting a tendency analysis unit, a comprehensive and dynamic evaluation of user satisfaction is achieved, which provides an important basis for the subsequent accurate marketing recommendations and personalized services. By setting a user feature determination unit to comprehensively construct a complete user feature, the recommendation system can more accurately provide personalized recommendations and services for each user, improve user satisfaction with the recommended content, and maintain the accuracy and effectiveness of the recommendation system.

[0040] It can be understood that the sensitivity analysis unit described in the embodiment of the present invention can calculate the user's price sensitivity by calculating the price elasticity coefficient, where the price elasticity coefficient = (percentage change in purchase quantity) / (percentage change in price). For example, when the price of the target product increases by 10% and the user's purchase quantity decreases by 20%, the price elasticity coefficient is -20% / 10%=-2. The absolute value of the price elasticity coefficient is compared with the preset absolute value to determine the user's price sensitivity based on the comparison result. When the absolute value of the price elasticity coefficient is greater than the preset absolute value, the user's price sensitivity is high, otherwise, the user's price sensitivity is low, where the preset absolute value is 1.

[0041] Specifically, the tendency analysis unit 42 includes: A rating acquisition subunit is used to acquire a number of historical purchase ratings of the user and determine the initial satisfaction level based on the historical purchase ratings; a vocabulary analysis subunit, connected to the score acquisition subunit, for acquiring evaluation content corresponding to any of the historical purchase scores, analyzing the evaluation content based on the sentiment analysis model, identifying and determining positive and negative words in the evaluation content according to the analysis results, respectively calculating the proportion of positive and negative words in the evaluation content, obtaining the positive proportion and the negative proportion, comparing the positive proportion with the negative proportion, and calculating a correction coefficient based on the comparison result; a calibration subunit, connected to the vocabulary analysis subunit, for calibrating the initial satisfaction level based on the correction coefficient to obtain an actual satisfaction level; A curve drawing subunit, connected to the calibration subunit, for drawing the satisfaction change graph based on a number of the actual satisfaction levels; The tendency analysis subunit is connected to the curve drawing subunit and is used to analyze the curve change trend corresponding to the satisfaction change graph and determine the user's purchase tendency degree according to the trend change analysis result.

[0042] It can be understood that the rating acquisition subunit described in the embodiment of the present invention can convert the rating into initial satisfaction through a simple linear transformation. For example, the user's 5 purchase ratings for a mobile phone product are obtained as 3 stars, 4 stars, 2 stars, 4 stars, and 3 stars respectively. The number of stars is converted into a value between 0 and 1. The formula is: initial satisfaction S0 = (number of stars - 1) / (5-1). For the above 5 ratings, the converted initial satisfactions are: 0.5, 0.75, 0.25, 0.75, and 0.5 respectively.

[0043] It can be understood that the sentiment analysis model described in the embodiment of the present invention can be selected by selecting the sentiment analysis model provided in some open source natural language processing libraries (such as NLTK, TextBlob, etc.), or by training a sentiment analysis model for a specific product or service field based on a deep learning framework (such as TensorFlow, PyTorch, etc.). The training data can be a large amount of collected evaluation texts with clear emotional tendency annotations (such as evaluation texts marked as positive, negative or neutral).

[0044] It can be understood that the vocabulary analysis subunit of the embodiment of the present invention is used to calculate the positive calibration coefficient based on the positive proportion and the negative proportion when the positive proportion is greater than the negative proportion, and to calculate the negative calibration coefficient based on the positive proportion and the negative proportion when the positive proportion is less than the negative proportion. When the positive proportion is equal to the negative proportion, the calibration coefficient is 1, wherein "positive" means that the calibration coefficient is used as an increasing factor to increase the initial satisfaction, and "negative" means that the calibration coefficient is used as a decreasing factor to decrease the initial satisfaction. For example, after removing stop words from a certain evaluation content, there are a total of 50 valid words, of which there are 20 positive words and 10 negative words, then the positive word is 1. The positive proportion is 0.4, and the negative proportion is 0.2. Comparing the positive proportion with the negative proportion, the positive proportion is greater than the negative proportion, then the positive correction coefficient is 0.4-0.2=0.2, then the calibration subunit determines the actual satisfaction = (1+0.2) × initial satisfaction. If a certain evaluation content has a total of 50 valid words after removing stop words, etc., including 10 positive words and 20 negative words, then the positive proportion is 0.2, and the negative proportion is 0.4. Comparing the positive proportion with the negative proportion, the positive proportion is less than the negative proportion, then the negative correction coefficient is 0.4-0.2=0.2, then the calibration subunit determines the actual satisfaction = (1-0.2) × initial satisfaction.

[0045] Specifically, the tendency analysis subunit includes: A tendency analysis block is used to identify the change trend of the curve, so as to determine the user's purchase tendency according to the trend change analysis result; The degree analysis block is connected to the tendency analysis block and is used to calculate the slope of the curve in the satisfaction change graph to determine the tendency degree according to the slope calculation result.

[0046] It can be understood that the satisfaction change graph described in the embodiment of the present invention takes time as the horizontal axis and actual satisfaction as the vertical axis, and presents the changes in user satisfaction with the target product in different time periods.

[0047] Specifically, the trend analysis block described in the embodiment of the present invention can identify the change pattern of user satisfaction over time by observing the curve trend in the satisfaction change graph. For example, if the curve shows a continuous upward trend, it indicates that the user's satisfaction with the product is gradually increasing, thereby reflecting that the user's purchase tendency for the product is increasing. On the contrary, if the curve shows a downward trend, it means that the user's satisfaction is decreasing, and the user's purchase tendency for the product may be weakening.

[0048] It can be understood that if the curve change trend is an upward trend, the degree analysis block described in the embodiment of the present invention calculates the upward slope, that is, the increase in satisfaction per unit time. For example, by selecting the coordinates of different time points on the curve, the slope calculation formula: slope = (satisfaction value at the next moment - satisfaction value at the previous moment) / (time at the next moment - time at the previous moment) is used to calculate several slope values, and the average of the several slope values ​​is used as the upward slope value, and the upward slope value is used as the quantitative result of the tendency degree. Similarly, if the curve change trend is a downward trend, the downward slope value is calculated, and the downward slope value is used as the quantitative result of the tendency degree. For a steady trend, the purchase tendency degree is 1.

[0049] It can be understood that the embodiment of the present invention determines the user's purchasing tendency degree based on the user's purchasing tendency and the tendency degree. For example, the purchasing tendency degree is compared with a preset degree. When the purchasing tendency degree is greater than the preset degree and the curve trend is rising, the user's purchasing tendency degree is determined to be.

[0050] It can be understood that the user feature determination unit described in the embodiment of the present invention is used to classify users according to the integrated user feature vector, for example, to divide users into different categories such as "high price sensitivity and high purchasing tendency group", "low price sensitivity and low purchasing tendency group", "high price sensitivity and low purchasing tendency group", and "low price sensitivity and high purchasing tendency group".

[0051] See also Figure 4 As shown, the association determination module 50 includes: A model building unit 51, used to build an initial correlation AI model based on a number of historical marketing feature information and a number of corresponding historical user feature information; A correlation determination unit 52, connected to the model building unit 51, for determining the correlation between the user feature and a plurality of the marketing features based on the initial correlation AI model; The priority determination unit 53 is connected to the relevance determination unit 52, and is used to determine the prediction priority based on the sorting results of the several relevance degrees.

[0052] Specifically, the embodiment of the present invention constructs an initial correlation AI model by setting a model construction unit, which lays a solid foundation for the subsequent accurate analysis of the correlation between user characteristics and marketing characteristics, so that the recommendation system can make more reasonable and accurate correlation judgments based on historical experience, which is helpful to improve marketing effectiveness. By setting a correlation determination unit based on the constructed initial correlation AI model, it is possible to accurately determine the correlation between user characteristics and marketing characteristics, which provides an important basis for the subsequent reasonable arrangement of the recommendation order of marketing plans, and helps to improve the accuracy of recommendations. By setting a priority determination unit, the recommendation priority of each marketing plan is clarified for the recommendation system, so that when the marketing plans are displayed to the user, they are arranged in order from high to low priority, ensuring that the user first sees the marketing plan that is most likely to meet his or her interests and needs, so that the recommended content is more in line with the user's actual needs, thereby improving the effectiveness and accuracy of the recommendation system.

[0053] It can be understood that the model building unit described in the embodiment of the present invention includes: for user characteristics, such as "high price sensitivity and high purchase tendency group", it is first necessary to encode it so that the model can process it. Assume that the price sensitivity "high" is encoded as 1 and "low" is encoded as 0; the purchase tendency degree is also encoded similarly, for example, "high" is 1, "medium" is 0.5, and "low" is 0, then the user characteristic of "high price sensitivity and high purchase tendency group" can be encoded in the form of a vector , for example (this is just a simple example, there may be more dimensions to fully represent user characteristics); For marketing features, take "high discount strategy" and "limited-time event" as examples. For "high discount strategy", based on the previous analysis of discount strategy, its discount rate and other related factors are quantified and coded. For example, if the discount rate is higher than a certain threshold (such as 20%), it is considered to be a "high discount strategy", and it is coded as For "high-time-limited activities", we encode them according to factors such as the duration of the activity. If the duration of an activity is shorter than a certain threshold (such as 3 days), it is considered a "high-time-limited activity" and can be encoded as , and then combine the encoding vectors of different marketing features to form a vector representing the comprehensive marketing features of the marketing plan, such as ; Determine the number of input layer nodes according to the dimensions of the user characteristics and marketing characteristics to be input. Assuming that the two user characteristics of user price sensitivity and purchase propensity are considered, as well as the two marketing characteristics of the marketing plan discount rate and activity duration, then the input layer needs to set up 4 nodes to receive these data respectively. The design of the hidden layer is relatively flexible. The number of hidden layers and the number of neurons in each layer can be determined based on experience and experiments. Generally speaking, you can try to set up 1-3 hidden layers at the beginning. The number of neurons in each layer can be selected between 1 / 2 and 2 times the number of input layer nodes. The number of nodes in the output layer is usually determined according to the output results. In the association model, the main thing is to output indicators that represent the strength of the correlation between user characteristics and marketing characteristics, so generally, one output layer node is sufficient; The encoded and preprocessed user feature vector and marketing feature vector are used as inputs to the initialized initial association AI model. The hidden layer of the model processes the input data according to the preset neuron connection structure and randomly initialized weight parameters. For example, if there are 6 neurons in a hidden layer, each node in the input layer will be connected to each neuron in the hidden layer through a certain weight. For example, the connection weight of the first element "1" of the user feature vector with the first neuron in the hidden layer may be 0.3, the connection weight with the second neuron may be 0.2, and the connection weight with the third neuron may be 0.5, etc. (these weights are randomly initialized and will be adjusted later). Similarly, each element of the marketing feature vector will also have a corresponding connection weight with the hidden layer neuron. When the data is input into the hidden layer, each neuron will perform a weighted sum operation based on the received input data and its connection weight, and then perform a nonlinear transformation through an activation function (such as ReLU, Sigmoid, etc.) to obtain the output of the hidden layer neuron. For example, after the first neuron in the hidden layer receives input from the user feature vector and the marketing feature vector, it performs weighted summation to obtain a value, assuming it is 2.5, and then processes it through the ReLU activation function (assuming that) If 2.5 is greater than 0, it outputs 2.5, otherwise it outputs 0. In this way, each neuron in the hidden layer will generate an output value to form a new vector, which is a feature transformation and processing result of the hidden layer on the input data; The output vector of the hidden layer will be used as the input of the output layer. The output layer calculates an indicator representing the strength of the association, such as the association score, based on its internal neuron connection structure and weight parameters (also determined through initialization and subsequent training). Assuming that the output layer has only one neuron, it will perform a weighted sum operation based on the received hidden layer output vector and its connection weight, and then process it through a suitable activation function (such as the Sigmoid function, which can map the output value to the range of 0-1 to facilitate the representation of the strength of the association) to obtain the final association score. For example, if after calculation and processing, the obtained association score is 0.8, this means that the correlation between the user feature and the marketing plan (with the corresponding marketing feature code) is strong.

[0054] It can be understood that the construction of the initial association AI model described in the embodiment of the present invention is a prior art. The above construction process is only to explain the working principle of the model construction unit in the embodiment of the present invention. A more complex and optimized model construction method can be adopted in actual application.

[0055] It can be understood that the prediction priority in the embodiment of the present invention can be determined from high to low according to several correlation degrees.

[0056] Specifically, the recommendation module 60 includes: A content determination unit, configured to use the marketing features corresponding to the target marketing plans as primary display content and the entire content of the target marketing plans as secondary display content; a page construction unit connected to the content determination unit, configured to determine a page area of ​​a user recommendation page, determine a plurality of target area areas and an arrangement order of a plurality of target marketing plans based on the predicted priority, and display a plurality of primary display contents on the user recommendation page based on the arrangement order and the plurality of target area areas; The real-time adjustment unit is connected to the page construction unit and is used to determine the target priority based on the real-time click count of the user's primary display content and the real-time browsing time and real-time purchasing behavior of the corresponding secondary display content.

[0057] It can be understood that the page construction unit described in the embodiment of the present invention is used to allocate a relatively larger target area for marketing plans with high predicted priorities, so as to more prominently display their contents and attract the user's attention. For example, if there are three marketing plans, the corresponding target areas are approximately 50% of the target area for the marketing plan with the highest priority, approximately 30% of the target area for the marketing plan with medium priority, and approximately 20% of the target area for the marketing plan with the lowest priority. The predicted priorities are Plan A, Plan B, and Plan C from high to low. Plan A may be allocated to a more conspicuous and larger area on the page, such as the upper half of the page, occupying a larger space; the target area of ​​Plan B may be relatively small and be placed in a suitable position after Plan A; the target area of ​​Plan C is even smaller and is placed in a further back position.

[0058] It can be understood that the real-time adjustment unit described in the embodiment of the present invention can determine the corresponding real-time evaluation value by calculating the weight sum of the real-time click count, real-time browsing time and real-time purchase behavior of any plan. For example, the weight of the real-time click count is set to 0.4, the weight of the real-time browsing time is set to 0.3, and the weight of the real-time purchase behavior is set to 0.3. Assuming that the real-time click count of Plan A is 10 times, the real-time browsing time is 30 seconds, and the user has a real-time purchase behavior (assuming that the score corresponding to the purchase behavior is 1, and the score corresponding to the non-purchase behavior is 0), then the real-time evaluation value of Plan A = real-time click count × click count weight coefficient + real-time browsing time × browsing time weight coefficient + real-time purchase behavior × purchase behavior weight coefficient = 10×0.4+30×0.3+1×0.3=13.3.

[0059] It can be understood that the target priority described in the embodiment of the present invention is determined by sorting the real-time evaluation values ​​corresponding to a number of target marketing plans from large to small.

[0060] Specifically, the feedback module includes: A comparison unit, used to compare the actual difference between the target priority and the predicted priority; A model adjustment unit is connected to the comparison unit and is used to adjust the initial association AI model based on the comparison result between the actual difference and the preset difference, or to reconstruct the association AI model to obtain a target association AI model.

[0061] It can be understood that the comparison unit described in the embodiment of the present invention determines the actual difference by judging the number of different rankings of the target marketing plans corresponding to the target priority and the predicted priority, so as to determine the actual difference based on the number of different rankings and the total number of marketing plans.

[0062] It can be understood that the model adjustment unit described in the embodiment of the present invention is used when the actual difference is less than the preset difference. If the initial association AI model is constructed based on a neural network, the parameters that may be adjusted include learning rate, regularization parameter, number of hidden layer neurons, etc. For example, if it is found that the model's prediction of the association between certain user features and marketing features is not accurate enough, the learning rate may be appropriately reduced so that the model can adjust the weights more finely during the training process to better capture the relationship between these features; or based on the analysis of the reasons for the differences, it is found that the model sets unreasonable feature weights for a certain type of marketing plan, such as the marketing feature of "gift value" is too low when predicting priority, resulting in inconsistency with the actual user behavior. At this time, the weight of the feature can be appropriately increased; for models built based on decision trees, the depth of the decision tree, the minimum number of leaf node samples, the selection criteria for dividing features, and other parameters can be adjusted. For example, if it is found that the decision tree model is too simplistic when predicting the priority of certain marketing plans, the depth of the decision tree may be appropriately increased to more finely divide the data set and better adapt to the complex relationship between user features and marketing features.

[0063] It can be understood that the model adjustment unit described in the embodiment of the present invention is also used to expand the training data when the actual difference is greater than the preset difference, and collect more historical data related to the target marketing plan and users, including data such as users' purchasing behavior, browsing behavior, and evaluation behavior in different time periods, as well as implementation effect data of different marketing plans in various scenarios. For example, if it is found that the model has a large deviation in predicting the priority of the marketing plan for a certain type of product, such as promotion activities for electronic products, more user behavior data and promotion activity case data related to electronic products are collected, and the expanded data set is re-preprocessed, including cleaning, encoding, normalization and other operations to make it meet the requirements of the model input, and then the re-preprocessed data set is used to retrain the initial associated AI model, and training is performed according to the selected loss function (such as mean square error, cross entropy, etc.) and optimization algorithm (such as stochastic gradient descent, Adam, etc.) until the performance indicators of the model on the validation set (such as accuracy, recall rate, mean square error, etc.) reach a satisfactory level or no longer have a significant improvement.

[0064] It can be understood that the model adjustment unit described in the embodiment of the present invention is also used to rebuild the associated AI model when the actual difference is extremely large and the existing model cannot meet the requirements through adjustment. According to business needs, data characteristics and problems with the previous model, a suitable machine learning algorithm or model architecture is reselected. For example, if the previous model performs poorly in processing nonlinear relationships, it may be considered to select a complex neural network architecture based on deep learning, such as a convolutional neural network (CNN) or a recurrent neural network (RNN) and its variants (such as LSTM, GRU, etc.); if the data has obvious classification characteristics and requires interpretability, an advanced variant of the decision tree may be selected, such as a random forest, a gradient boosting tree, etc. According to the newly selected model architecture, a series of construction processes such as data collection, preprocessing, model design (including determining the input layer, hidden layer, output layer, etc.) and training are performed, which is similar to the process of building the initial associated AI model described above, but it will fully learn from the lessons of the previous model and carry out targeted design and optimization for the problems that have occurred before.

[0065] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0066] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An intelligent recommendation system integrating big data and AI, characterized in that: include: A program acquisition module is used to obtain several target marketing programs for any target product; A scheme analysis module, connected to the scheme acquisition module, for analyzing a number of the target marketing schemes, obtaining a number of keywords and a number of data indicators, and determining a number of marketing features based on the keywords and the data indicators; A data acquisition module is used to collect some historical behavior data of users based on a preset period; a data analysis module connected to the data acquisition module, configured to analyze the historical behavior data, obtain the user's price sensitivity based on the analysis results of the purchase records, determine the user's purchase tendency based on the evaluation records, and determine the user's characteristics based on the user's price sensitivity and the user's purchase tendency; A correlation determination module, connected to the solution analysis module and the data analysis module, for analyzing the correlation between the user characteristics and the marketing characteristics based on the initial correlation AI model, and determining the prediction priority of the target marketing solutions based on the correlation analysis results; A recommendation module, connected to the association determination module, is used to determine the marketing plan page display content, build a user recommendation page based on the predicted priority and the page display content, and collect real-time behavior data of users when browsing the recommendation page, and determine the target priority of several target marketing plans based on the real-time user behavior data; A feedback module is connected to the recommendation module and is used to adjust the initial association AI model based on the comparison result between the predicted priority and the target priority to obtain a target association AI model.

2. The intelligent recommendation system integrating big data and AI according to claim 1, characterized in that: The solution analysis module includes: A keyword determination unit, configured to perform word segmentation processing on any of the target marketing plans to obtain a number of actual word segments, and determine a number of keywords based on word frequency statistics of the number of actual word segments and matching results between the number of actual word segments and a preset marketing vocabulary; A data indicator determination unit, used to extract a number of actual data from any of the target marketing plans, determine the data types corresponding to the number of actual data, determine the corresponding data indicators according to the data types, and obtain a number of data indicators; The marketing feature determination unit is connected to the keyword determination unit and the data indicator determination unit, and is used to determine the marketing feature based on a plurality of the keywords and a plurality of the data indicators.

3. The intelligent recommendation system integrating big data and AI according to claim 2 is characterized in that: The data indicator determination unit includes: A type determination subunit, used to determine the data type based on a data format corresponding to any of the actual data and a connecting word connected thereto; The indicator determination subunit is connected to the type determination subunit and is used to compare the data type with a preset data type in a preset indicator calculation library, perform calculation based on the comparison result, and obtain a data indicator.

4. The intelligent recommendation system integrating big data and AI according to claim 3 is characterized in that: The data acquisition module comprises: A recording unit is used to record a number of historical data corresponding to the user on the product platform; The determination unit is connected to the recording unit and is used to extract a plurality of purchase records and a plurality of evaluation records from a plurality of historical data within a preset period.

5. The intelligent recommendation system integrating big data and AI according to claim 4 is characterized in that: The data analysis module includes: a sensitivity analysis unit, configured to determine the purchase quantity of the target product and the corresponding purchase amount of the target product purchased by the user based on the purchase records, and determine the user's price sensitivity based on the purchase quantity and the corresponding purchase amount; a tendency analysis unit, for determining a number of actual satisfactions of the user to the target product based on the number of evaluation records, drawing a satisfaction change graph based on the number of actual satisfactions, analyzing the satisfaction change graph, and determining the user's purchase tendency according to the analysis result; The user feature determination unit is connected to the sensitivity analysis unit and the tendency analysis unit, and is used to determine the user feature according to the user's price sensitivity and the user's purchase tendency.

6. The intelligent recommendation system integrating big data and AI according to claim 5 is characterized in that: The tendency analysis unit comprises: A rating acquisition subunit is used to acquire a number of historical purchase ratings of the user and determine the initial satisfaction level based on the historical purchase ratings; a vocabulary analysis subunit, connected to the score acquisition subunit, for acquiring evaluation content corresponding to any of the historical purchase scores, analyzing the evaluation content based on the sentiment analysis model, identifying and determining positive and negative words in the evaluation content according to the analysis results, respectively calculating the proportion of positive and negative words in the evaluation content, obtaining the positive proportion and the negative proportion, comparing the positive proportion with the negative proportion, and calculating a correction coefficient based on the comparison result; a calibration subunit, connected to the vocabulary analysis subunit, for calibrating the initial satisfaction level based on the correction coefficient to obtain an actual satisfaction level; A curve drawing subunit, connected to the calibration subunit, for drawing the satisfaction change graph based on a number of the actual satisfaction levels; The tendency analysis subunit is connected to the curve drawing subunit and is used to analyze the curve change trend corresponding to the satisfaction change graph and determine the user's purchase tendency degree according to the trend change analysis result.

7. The intelligent recommendation system integrating big data and AI according to claim 6 is characterized in that: The tendency analysis subunit comprises: A tendency analysis block is used to identify the change trend of the curve, so as to determine the user's purchase tendency according to the trend change analysis result; The degree analysis block is connected to the tendency analysis block and is used to calculate the slope of the curve in the satisfaction change graph to determine the tendency degree according to the slope calculation result.

8. The intelligent recommendation system integrating big data and AI according to claim 7, characterized in that: The association determination module comprises: A model building unit, used to build an initial correlation AI model based on a number of historical marketing feature information and a number of corresponding historical user feature information; A correlation determination unit, connected to the model building unit, for determining the correlation between the user feature and a plurality of the marketing features based on the initial correlation AI model; The priority determination unit is connected to the relevance determination unit and is used to determine the prediction priority based on the sorting results of the several relevance degrees.

9. The intelligent recommendation system integrating big data and AI according to claim 8, characterized in that: The recommendation module includes: A content determination unit, configured to use the marketing features corresponding to the target marketing plans as primary display content and the entire content of the target marketing plans as secondary display content; a page construction unit connected to the content determination unit, configured to determine a page area of ​​a user recommendation page, determine a plurality of target area areas and an arrangement order of a plurality of target marketing plans based on the predicted priority, and display a plurality of primary display contents on the user recommendation page based on the arrangement order and the plurality of target area areas; The real-time adjustment unit is connected to the page construction unit and is used to determine the target priority based on the real-time click count of the user's primary display content and the real-time browsing time and real-time purchasing behavior of the corresponding secondary display content.

10. The intelligent recommendation system integrating big data and AI according to claim 9, characterized in that: The feedback module comprises: A comparison unit, used to compare the actual difference between the target priority and the predicted priority; A model adjustment unit is connected to the comparison unit and is used to adjust the initial association AI model based on the comparison result between the actual difference and the preset difference, or to reconstruct the association AI model to obtain a target association AI model.

Images