Personalized analysis method for cultural and creative products based on collaborative filtering

Abstract

The invention relates to the technical field of electronic commerce, in particular to a personalized analysis method for cultural and creative products based on collaborative filtering, which comprises the following steps: acquiring browsing, collection and purchase behavior data of a user, calculating interval change of adjacent behaviors, and adjusting behavior influence based on browsing frequency, collection density and purchase conversion. And calculating a behavior contribution proportion in combination with the collection frequency and the stay duration, and extracting a short-term behavior trend to obtain a user behavior change trend parameter. According to the method, the behavior change rate and the time interval are combined for screening, the recommended content can quickly adapt to user interest changes, the recommendation matching degree is improved through utilization of social interaction data, social recommendation is more targeted, the scene applicability of the recommended content is enhanced through consideration of environmental factors such as geographic positions and weather, and the user experience is improved. The dynamic calculation of the multi-dimensional weight optimizes the recommendation sequence, and improves the long-term effectiveness and user experience of the recommendation system.

Description

Technical Field

[0001] The present invention relates to the field of e-commerce technology, and in particular to a personalized analysis method for cultural and creative products based on collaborative filtering. Background Art

[0002] The field of e-commerce encompasses internet-based information exchange, commercial transactions, and personalized recommendations. Its core focus is optimizing product sales and service experiences through technologies such as data mining, user behavior analysis, and personalized recommendations. E-commerce systems typically leverage machine learning models and data analysis methods to model user preferences and use intelligent algorithms to provide customized product or service recommendations. Collaborative filtering, as a key personalized recommendation technology, is widely used in scenarios such as product recommendations, content distribution, and precision marketing to enhance user experience and boost transaction conversions.

[0003] The collaborative filtering-based personalized analysis method for cultural and creative products utilizes collaborative filtering technology to analyze historical user behavior data to extract personalized features and generate recommendations. This method encompasses specific technical aspects such as user behavior data collection, similarity calculation, and user interest prediction. First, user behavior data on browsing, collecting, and purchasing cultural and creative products is collected, and a user-item interaction matrix is ​​constructed. Second, user-based or item-based collaborative filtering is used to calculate similarities between users or items, employing specific mathematical methods to calculate similarity. Based on this calculated similarity data, cultural and creative products that may be of interest to the user are inferred, and recommendations are output according to specific ranking rules.

[0004] The existing technologies have problems in the recommendation mechanism, such as insufficient sensitivity in capturing changes in interests, insufficient timeliness of recommended content, low matching degree of social recommendations, limited environmental adaptation capabilities, and delayed adjustment of recommendation order. Failure to conduct detailed analysis of changes in user behavior intervals makes it difficult to accurately judge the trend of interest deviation, making it impossible for recommended content to respond to changes in interest in a timely manner. Social recommendations are only based on static similarity calculations of collaborative filtering, ignoring the interaction intensity and common attention trends between users, resulting in low relevance of recommended content, which affects user acceptance. In terms of environmental adaptation, the dynamic changes in the user's scenario, such as the impact of weather and time on demand, are not fully considered, resulting in a lower matching degree of recommended content in different situations. The recommendation sorting method is difficult to adjust with real-time changes in user interaction data, making it difficult to continuously optimize recommended content, affecting the user's long-term usage experience. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the existing technology and propose a personalized analysis method for cultural and creative products based on collaborative filtering.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for personalized analysis of cultural and creative products based on collaborative filtering, comprising the following steps: S1: Obtain user browsing, collection, and purchase behavior data, calculate the changes in adjacent behavior intervals, adjust the behavior impact based on browsing frequency, collection density, and purchase conversion, calculate the behavior contribution ratio based on collection frequency and stay time, extract short-term behavior trends, and obtain user behavior change trend parameters; S2: Based on the user behavior change trend parameters, calculate the recent behavior change amplitude, filter out categories with obvious fluctuations, extract key interest transfer points, calculate the matching weight of individual user behavior and group interest, adjust the recommendation ranking, and obtain the user interest shift weight; S3: Obtain user social interaction data, calculate the interest fit between the user and their friends, select friend groups with high interaction and similar interests, calculate the shared interest matching index, adjust the recommendation weight based on the friend group's attention intensity for cultural and creative products, and optimize the recommendation based on the social hierarchy to obtain the proportion of social related recommendations; S4: Obtain the user's geographic location, access time, weather, and device type, calculate the scene matching degree, adjust the recommendation category based on the user's location, modify the recommendation structure based on time, screen suitable products, dynamically adjust the matching weight, and obtain the environment adaptation adjustment coefficient.

[0007] As a further solution of the present invention, the user behavior change trend parameters include time interval change value, behavior impact weight, category contribution ratio, and short-term trend index; the user interest deviation weight includes behavior change amplitude, interest transfer point, individual matching weight, and recommendation ranking weight; the social association recommendation ratio includes interest fitting degree, shared interest matching degree, friend attention intensity, and social interaction level weight; the environmental adaptation adjustment coefficient includes scene matching degree, recommendation category adjustment coefficient, display structure correction parameter, and product screening weight.

[0008] As a further solution of the present invention, the specific steps of obtaining user browsing, collection, and purchase behavior data, calculating the change in adjacent behavior intervals, adjusting the behavior impact based on browsing frequency, collection density, and purchase conversion, calculating the behavior contribution ratio based on collection frequency and stay time, extracting short-term behavior trends, and obtaining user behavior change trend parameters are as follows: S101: Obtain the user's browsing, collection, and purchase behavior data, arrange them in chronological order, calculate the time intervals between adjacent behaviors, call the time points of the previous and next behaviors to calculate the time difference, use the time interval mean of the behavior sequence as a benchmark, determine the change of each time interval, and generate the behavior time interval change coefficient; S102: Calculate the influence of the behavior category on the overall behavior sequence by calling the behavior time interval variation coefficient and combining it with browsing frequency, collection density, and purchase conversion. Increase the influence when the interval is shortened and decrease the influence when the interval is increased. Calculate the contribution ratio based on the occurrence ratio of the behavior category and the time interval variation trend to generate the behavior category influence ratio value. S103: Call the behavior category influence weight value, combine it with the collection frequency and page dwell time, analyze the short-term changes of the behavior category, call the behavior category influence weight value of adjacent time periods, calculate the increase and decrease rate, extract the short-term behavior change trend, and generate user behavior change trend parameters.

[0009] As a further solution of the present invention, based on the user behavior change trend parameter, the recent behavior change amplitude is calculated, the categories with obvious fluctuations are screened, key interest transfer points are extracted, the matching weight of individual user behavior and group interest is calculated, and the recommendation ranking is adjusted to obtain the user interest shift weight. The specific steps are: S201: Obtain the user behavior change trend parameter, extract the recent behavior change rate, set the behavior time interval, calculate the change range of each category of behavior, call the user's behavior record data, calculate the increase or decrease range based on the change amount of each category of behavior within the time interval, and obtain the behavior category change range; S202: Calculating the change range of the behavior categories, screening the behavior categories with obvious fluctuations based on the number of favorites, browsing paths, and purchase times, calculating the fluctuation range of each category of behavior within a set time interval, calculating the category change ratio, extracting the behavior categories whose change rate exceeds the set threshold, identifying the critical time points of the behavior category changes, and extracting the key interest transfer points; S203: Call the key interest transfer point, calculate the matching weight of individual user behavior and group interest, adjust the priority of recommended content based on the changing trend of user behavior category and analyze the degree of deviation from group interest, and obtain the user interest deviation weight.

[0010] As a further solution of the present invention, the formula for calculating the matching weight between individual user behavior and group interests is specifically: ; in, Represents the matching weight between individual user behavior and group interests, Represents the user in the category The frequency of behaviors on Representative groups in categories The average frequency of behaviors on Represents the user in the category The weight of the behavioral change trend, Represents the user in the time period The behavioral deviation value within Represents the normalized adjustment factor for matching all categories of user behavior. Represents the total number of user behavior categories, Represents the total number of user behavior time periods.

[0011] As a further solution of the present invention, the specific steps of obtaining user social interaction data, calculating the interest fit between the user and friends, screening friend groups with high interaction and similar interests, calculating the shared interest matching index, adjusting the recommendation weight according to the friend group's attention intensity to cultural and creative products, and combining the social level optimization recommendation to obtain the social association recommendation ratio are as follows: S301: Obtain the user's social interaction data, extract the browsing, collection, and purchase records of friends, call the user and friends' behavior data, calculate the matching degree based on the number of browsing, collection, and purchase behaviors of the same category, and obtain the interest fit between the user and friends; S302: Calling the interest fit between the user and their friends, screening friend groups with high interaction frequency and similar interests, calling the interaction records between the user and each friend, recording the number of interactions and duration, screening friend groups with interaction times exceeding a set threshold, calculating the matching degree index of shared interests within the group, analyzing the overall interest shift trend based on the intensity of each friend group's attention to cultural and creative products, and extracting the shared interest matching degree index; S303: Calling the shared interest matching degree index, combining the user's social interaction level, adjusting the influence proportion of the recommended product, adjusting the recommendation order according to the degree of association between the user and friends of different levels, and obtaining the social association recommendation proportion.

[0012] As a further solution of the present invention, the calculation formula of the matching degree index of shared interests within the group is specifically: ; in, An indicator of the degree of matching that represents shared interests within the group, Representing users and friends The number of interactions, Representing users and friends Matching degree in the same interest category, Represents an individual in a friend group The intensity of attention paid to cultural and creative product categories, Represents the normalized adjustment factor of overall user social interaction, Represents the total number of friends who meet the interaction frequency threshold, Represents the total number of people in the user's friend group.

[0013] As a further solution of the present invention, the specific steps of obtaining the user's geographic location, access time, weather, and device type, calculating the scene matching degree, adjusting the recommended categories based on the user's location, correcting the recommendation structure based on time, screening suitable products, and dynamically adjusting the matching weight to obtain the environment adaptation adjustment coefficient are as follows: S401: Obtain the user's geographic location, access time, weather conditions, and device type, call the latitude and longitude data of the user's location, access time, current weather type, and device model, calculate the feature matching degree of the user's scene, and obtain the user scene feature matching value; S402: Utilize the user scenario feature matching value, adjust the recommended categories of cultural and creative products based on the user's location, utilize the category tag of the user's current location, filter cultural and creative products with high correlation with the category tag, modify the display structure of recommended content based on the visit time period, utilize the product click preference values ​​for differentiated time periods, filter high-preference categories and optimize the display order, filter suitable products based on weather conditions, extract weather condition and product matching factors, adjust the display ratio in the recommendation list, and obtain scenario adaptation adjustment weights; S403: Call the scene adaptation adjustment weight, dynamically adjust the scene matching weight, combine the user access scene characteristics and the recommendation adjustment coefficient, calculate the adjustment proportion of the environment adaptation recommendation, and obtain the environment adaptation adjustment coefficient.

[0014] As a further embodiment of the present invention, the method further includes: S5: calculating a recommendation ranking weight based on the user interest offset weight, the social association recommendation ratio, and the environment adaptation adjustment coefficient, screening products with a high recent interaction frequency and increasing their priority; lowering the ranking of similar products if their popularity decreases; adjusting the display strategy in combination with continuous interaction, optimizing the ranking order, and obtaining a product personalized analysis solution; The product personalization analysis solution includes recommendation ranking weight, interaction frequency coefficient, attention adjustment parameter, and display optimization strategy.

[0015] As a further solution of the present invention, based on the user interest offset weight, the social association recommendation ratio and the environment adaptation adjustment coefficient, the recommendation ranking weight is calculated, and products with high recent interaction frequency are screened for priority. If the attention of similar products decreases, the ranking is lowered. Combined with continuous interaction, the display strategy is adjusted to optimize the arrangement order. The specific steps of obtaining a product personalized analysis solution are as follows: S501: Obtain the user interest offset weight, the social association recommendation ratio, and the environment adaptation adjustment coefficient, call each weight parameter, calculate the ranking weight of the recommended content, calculate the ranking score based on the influence ratio of each weight parameter, and obtain the recommendation ranking weight; S502: Utilize the recommendation ranking weights to filter products with high recent interaction frequencies and increase their recommendation priority. Utilize the user's recent browsing, collection, and purchase records to calculate the interaction frequency of each product. Filter high-frequency products based on the interaction frequency threshold. If the attention of each product decreases, utilize the access trend to calculate the decrease rate and adjust the ranking score. Calculate the interaction stability based on the user's continuous interaction status, modify the recommendation display strategy, and obtain an optimized recommendation ranking result. S503: Call the optimization recommendation ranking result, adjust the order of recommended content according to product priority, calculate content display weight based on user interaction preferences, optimize the display structure, and obtain a product personalized analysis plan.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by combining the behavior change rate and time interval screening, the recommended content can quickly adapt to changes in user interests. The use of social interaction data improves the recommendation matching degree, making social recommendations more targeted. The consideration of environmental factors such as geographic location and weather enhances the scenario applicability of the recommended content. The dynamic calculation of multi-dimensional weights optimizes the recommendation order and improves the long-term effectiveness of the recommendation system and user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 It is a schematic diagram of the steps of the present invention. DETAILED DESCRIPTION

[0019] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0020] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as an "exemplary" in the present invention should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of the word "exemplary" is intended to present concepts in a concrete manner. Furthermore, in the embodiments of the present invention, "and / or" can mean both or either of the two.

[0021] In the embodiments of the present invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, when the distinction is not emphasized, the meanings they convey are the same. The terms "of," "corresponding," and "corresponding" may sometimes be used interchangeably. It should be noted that, when the distinction is not emphasized, the meanings they convey are the same.

[0022] In the embodiments of the present invention, sometimes a subscript such as W1 may be written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are the same.

[0023] In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0024] See also Figure 1 , a personalized analysis method for cultural and creative products based on collaborative filtering, comprising the following steps: S1: Obtain user browsing, collection, and purchase behavior data, arrange them in chronological order, extract the time intervals between adjacent behaviors, calculate the change in time intervals, and adjust the impact of behaviors based on browsing frequency, collection density, and purchase conversion. If the interval between behaviors shortens, the impact is increased, while if the interval between behaviors increases, the impact is reduced. The contribution ratio of behavior categories is calculated based on collection frequency and page dwell time, and short-term behavior change trends are extracted to obtain user behavior change trend parameters. S2: Based on the user behavior change trend parameters, extract the change rate of recent behavior, set the behavior time interval, calculate the magnitude of behavior change in each category, filter out behavior categories with significant fluctuations based on the number of collections, browsing paths, and purchase times, extract key interest transfer points, calculate the matching weight of individual user behavior and group interests, adjust the priority of recommended content, and obtain the user interest shift weight; S3: Obtain the user's social interaction data, extract friends' browsing, collection, and purchase records, calculate the interest fit between the user and friends, screen friend groups with high interaction frequency and similar interests, calculate the matching degree index of shared interests, adjust the influence ratio of recommended products based on the friend group's recent attention to cultural and creative products, and adjust the recommendation order based on the user's social interaction level to obtain the proportion of social association recommendations; S4: Obtain the user's geographic location, visit time, weather conditions, and device type, calculate the feature matching degree of the user's scene, adjust the recommended categories of cultural and creative products based on the user's location, modify the display structure of recommended content based on the time period, select suitable products based on weather conditions, dynamically adjust the scene matching weight, and obtain the environment adaptation adjustment coefficient; S5: Based on the user interest offset weight, the proportion of social association recommendations, and the environmental adaptation adjustment coefficient, the ranking weight of the recommended content is calculated. Products with high recent interaction frequency are screened to increase the recommendation priority. If the attention of similar products decreases, the recommendation ranking is lowered. The display strategy is adjusted based on the user's continuous interaction situation, the content arrangement order is optimized, and a product personalized analysis plan is obtained.

[0025] User behavior change trend parameters include time interval change value, behavior impact weight, category contribution ratio, and short-term trend indicators; user interest deviation weight includes behavior change amplitude, interest transfer point, individual matching weight, and recommendation ranking weight; social association recommendation ratio includes interest fitting degree, shared interest matching degree, friend attention intensity, and social interaction level weight; environmental adaptation adjustment coefficient includes scene matching degree, recommendation category adjustment coefficient, display structure correction parameter, and product screening weight; product personalization analysis plan includes recommendation ranking weight, interaction frequency coefficient, attention adjustment parameter, and display optimization strategy.

[0026] The specific steps of S1 are: S101: Obtain the user's browsing, collection, and purchase behavior data, arrange them in chronological order, calculate the time intervals between adjacent behaviors, call the time points of the previous and next behaviors to calculate the time difference, use the time interval mean of the behavior sequence as a benchmark, determine the change of each time interval, and generate the behavior time interval change coefficient; Obtain the user's browsing, collection, and purchase behavior data, including each user's browsing history, collection history, and purchase history, and arrange them in chronological order. After sorting each behavior according to the time axis, extract the timestamp information of adjacent behaviors, call the time points of the previous and next behaviors to calculate the time difference, that is, for each group of adjacent behaviors, calculate the time interval, define the time interval as the timestamp of the latter behavior minus the timestamp of the previous behavior, store all adjacent behavior time interval data, calculate the mean of all time intervals in turn as the benchmark value, use the mean as the judgment standard, compare all time interval values ​​with the mean, and calculate the variation coefficient of each time interval. The calculation method is to divide the difference between each interval value and the mean by the mean, that is, variation coefficient = (current time) If the coefficient of variation is negative, it means that the current time interval is shorter than the mean. If it is positive, it means that the current time interval is longer than the mean. Suppose a user's browsing behavior records are as follows: T1=12:00, T2=12:05, T3=12:15, T4=12:30, then the adjacent time intervals are 5 minutes, 10 minutes, and 15 minutes respectively. The mean is (5+10+15) / 3=10 minutes, then the coefficient of variation of each time interval is (5-10) / 10=-0.5, (10-10) / 10=0, and (15-10) / 10=0.5. All the coefficients of variation are stored and sorted for subsequent calculation of the impact proportion of the behavior category.

[0027] S102: The coefficient of variation of the time interval between behaviors is called, and the influence of the behavior category on the overall behavior sequence is calculated by combining browsing frequency, collection density, and purchase conversion. The influence is increased when the interval is shortened, and decreased when the interval is increased. The contribution ratio is calculated based on the occurrence ratio of the behavior category and the trend of time interval variation, and the behavior category influence ratio value is generated. Combined with the user's browsing frequency, collection density, and purchase conversion, the number of users' browsing times, collection times, and purchase times within a period of time are extracted respectively, and the degree of influence of the behavior category on the overall behavior sequence is calculated. For each user, the number of occurrences of their behavior category is counted and classified into three categories: browsing, collection, and purchase. The proportion of each behavior category is calculated, that is, the proportion of behavior category = the number of occurrences of a certain behavior category / the total number of behaviors. Combined with the trend of time interval changes, its impact on the behavior pattern is analyzed. In the case of shortened time intervals, the influence proportion of the behavior category is increased. For example, if the browsing behavior interval change coefficient of a user is negative three times in a row, then the The influence ratio of the behavior category should be increased, and for the case of interval growth, its influence ratio should be reduced. The calculation method of influence ratio is: influence ratio = behavior category ratio × (1 + variation coefficient). For example, the browsing behavior category ratio of a user is 0.4, and the average value of the time interval variation coefficient is -0.2, then its influence ratio is calculated as 0.4×(1+(-0.2))=0.32. At the same time, the contribution ratio is calculated to reflect the contribution of each behavior category in the overall behavior sequence. The contribution ratio is calculated as: contribution ratio = influence ratio / (browsing influence ratio + collection influence ratio + purchase influence ratio), and finally the behavior category influence ratio value is obtained.

[0028] S103: Analyze the short-term changes in behavior categories by calling the influence weight of behavior categories, combining the frequency of favorites and the duration of stay on the page, and call the influence weight of behavior categories in adjacent time periods to calculate the increase and decrease rates, extract the short-term behavior change trend, and generate user behavior change trend parameters. Combined with users' favorite frequency and page dwell time, we analyze short-term changes in user behavior categories. We count the number of favorites and the time interval between favorites in adjacent time periods and calculate favorite frequency = number of favorites / duration. We also record the duration of users on each page. The dwell time is calculated as the exit time minus the entry time. We also calculate the influence proportion of users' behavior categories in different time periods and calculate the increase / decrease rate, that is, the rate of change of the influence proportion of the behavior category = (influence proportion of the current time period - influence proportion of the previous time period) / influence proportion of the previous time period. For example, if a user's browsing influence proportion in the previous time period is 0.32 and the current time period is 0.40, then the increase / decrease rate = (0.40-0.32) / 0.32 = 0.25. A positive increase / decrease rate indicates that the behavior category has strengthened in the short term, while a negative increase / decrease rate indicates that the behavior category has weakened in the short term. Based on the trend of the increase / decrease rate, we extract short-term behavior change trend parameters and store them in the user behavior trend data for further analysis and optimization.

[0029] The specific steps of S2 are: S201: Obtain user behavior change trend parameters, extract the change rate of recent behavior, set the behavior time interval, calculate the change range of each category of behavior, call the user's behavior record data, calculate the increase or decrease range based on the change amount of each category of behavior within the time interval, and obtain the change range of the behavior category; Extract the rate of change of recent behaviors, set the behavior time interval, divide the time window based on the time axis, call the user's behavior record data, count the behaviors of each category within the set time interval, and calculate the change of each category of behavior, that is, count the total number of behaviors of this category at the starting point and end point of the time interval respectively. The change is calculated by subtracting the starting statistical value from the ending statistical value, and calculating the increase or decrease, that is, the ratio of the change to the length of the time interval. If the change of the behavior category is positive, it means that the behavior has increased within the set time interval. If it is negative, it means that the behavior has increased. To reduce this, the calculation method for the behavior change amplitude is set as follows: behavior change amplitude = change amount / time interval length. For example, a user browses 50 times in the first 7 days and 80 times in the next 7 days, then the browsing behavior change amount is 80-50=30. The time interval length is 7 days, then the browsing behavior change amplitude is 30 / 7=4.29, indicating that the user's browsing behavior increases by 4.29 times per day. If this value is negative, it means that the behavior decreases. The change amplitudes of all categories are stored for subsequent analysis of the fluctuation of behavior categories.

[0030] S202: Calculate the change range of behavior categories, select behavior categories with obvious fluctuations based on the number of favorites, browsing paths, and purchase times, calculate the fluctuation range of each category within a set time interval, calculate the category change ratio, extract behavior categories with a change rate exceeding a set threshold, identify the critical time points of behavior category changes, and extract key interest transfer points; Based on the number of collections, browsing paths, and purchase times, the behavior categories with obvious fluctuations are screened out, the fluctuation standards of the behavior categories are set, the fluctuation amplitudes of the behavior categories are sorted by numerical value, and the first several behavior categories with the largest fluctuation amplitudes are extracted as the categories with obvious fluctuations. The fluctuation amplitude of each category behavior within the set time interval is called, and the change ratio of the category is calculated, that is, the ratio of the fluctuation amplitude to the average change amplitude. The calculation method of the set change ratio is: change ratio = change amplitude of the behavior category / mean of the change of the behavior category. For example, within a certain time interval, the change amplitude of the browsing behavior is 4.29, the change amplitude of the collection behavior is 1.5, and the change amplitude of the purchase behavior is 0.5, then the behavior change average is 0.5. The value is calculated as (4.29+1.5+0.5) / 3=2.1, so the change ratio of browsing behavior is 4.29 / 2.1=2.04, the change ratio of collection behavior is 1.5 / 2.1=0.71, and the change ratio of purchase behavior is 0.5 / 2.1=0.24. The behavior categories with change ratios exceeding the set threshold are extracted. The threshold is set to 1.5 times the mean value of behavior change, that is, 2.1×1.5=3.15. The change ratio of browsing behavior exceeds the threshold, and the behavior category is extracted. The critical time point of the behavior change of this category is identified, that is, the inflection point of the statistical behavior change amplitude on the timeline is extracted, and the key interest transfer points are extracted for the subsequent calculation of the matching weight of user behavior and group interest.

[0031] S203: Calling key interest transfer points, calculating the matching weight between individual user behavior and group interest, adjusting the priority of recommended content based on the changing trend of user behavior categories and analyzing the degree of deviation from group interest, and obtaining the user interest deviation weight; The formula for calculating the matching weight between individual user behavior and group interests is as follows: ; in, Represents the matching weight between individual user behavior and group interests, Represents the user in the category The frequency of behaviors on Representative groups in categories The average frequency of behaviors on Represents the user in the category The weight of the behavioral change trend, Represents the user in the time period The behavioral deviation value within Represents the normalized adjustment factor for matching all categories of user behavior. Represents the total number of user behavior categories, The total number of time periods representing user behavior: The formula is used to calculate the matching weight between individual user behavior and group interests , reflecting the degree of difference between individual user behavior and group interests. This calculation considers the difference between the frequency of user behavior in different categories and the group average frequency, adjusts it by the weight of the behavior change trend of each category, and finally standardizes it by the deviation value of user behavior and a normalization adjustment factor.

[0032] parameter (User in category Frequency of behaviors above (%): Actual acquisition: For example, through the data monitoring system, users’ specific categories of data within a certain period of time are collected. (such as cultural products) and divide it by the total number of behaviors of the user during that time.

[0033] Assuming that the number of user behaviors on cultural products is 120 and the total number of behaviors is 1000, then .

[0034] parameter (Groups in category Average frequency of behaviors on (%): Actual acquisition: Collect similar users in the same category through the data analysis system The average behavior frequency ratio.

[0035] Assuming that the average frequency of group behavior on cultural products is 0.15, then .

[0036] parameter (User in category Behavioral change trend weight on (weight): Setting basis: According to the user in the category The trend of behavior frequency changes and the standard deviation of past behaviors are used to quantify the trend, and the trend weight reflects the significance of the behavior change.

[0037] Assuming that based on past data, the weight of the behavior change trend of cultural products is 1.2, then .

[0038] parameter (Users in time period Behavioral deviation value within): Actual acquisition: Calculate the standard deviation between the user's behavior frequency and the average behavior frequency within a certain time period.

[0039] Assume that in the past week, the user's behavior deviation value is calculated to be 0.05, then .

[0040] parameter (Normalization adjustment factor for all categories of user behavior matching): Basis: To ensure the balance of the calculation results, a normalization factor is introduced to adjust the scale of the denominator.

[0041] Assuming the normalization factor is set to 1, then .

[0042] Calculation process: Calculate the absolute difference for each class and multiply by the weight: ; Calculate the cumulative sum of all categories (assuming there is only one category): ; Calculate the sum of squared behavioral deviations and normalize the adjustment in the denominator: ; Finally, the matching weight between individual user behavior and group interests is calculated: ; The results show that there is a slight offset between user interests and group interests, which is reflected in the fact that users' attention to cultural products is lower than the group average. This offset weight can be used to adjust the content priority in the recommendation system to make the recommendations more closely aligned with the user's actual interests.

[0043] The specific steps of S3 are: S301: Obtain the user's social interaction data, extract the browsing, collection, and purchase records of friends, call the user and friends' behavior data, calculate the matching degree based on the number of browsing, collection, and purchase behaviors of the same category, and obtain the interest fit between the user and friends; Extract friends' browsing, collection, and purchase records, call the user and friends' behavior data, classify them according to behavior categories, calculate the number of behaviors of users and friends in each category, count the number of browsing, collection, and purchase behaviors in the same category, and calculate the degree of matching between users and friends in the same category. The matching degree is calculated as follows: matching degree = 1-|(user behavior category number - friend behavior category number) / (user behavior category number + friend behavior category number)|. For example, if a user browses 60 times in a week and his friend browses 50 times, the browsing behavior matching degree is calculated as 1-|(60-50) / (60+50)|=1-|10 / 110|=0.91. A matching degree close to 1 indicates that the user and their friends have similar behavioral interests. A lower matching degree indicates that the user and their friends have significantly different interests. The same calculation method is used for collection and purchase behaviors. After obtaining the matching degrees of all behavior categories, the overall interest fit between the user and their friends is calculated. The overall interest fit is calculated as the weighted average of all behavior matching degrees. The weight of each behavior category is set. For example, the browsing behavior weight is set to 0.5, the collection behavior weight is set to 0.3, and the purchase behavior weight is set to 0.2. The overall interest fit between the user and their friends is calculated as 0.5×browsing matching degree + 0.3×collection matching degree + 0.2×purchase matching degree, and the interest fit between the user and their friends is finally obtained.

[0044] S302: Query the interest compatibility between the user and their friends, select friend groups with high interaction frequency and similar interests, retrieve the interaction records between the user and each friend, record the number of interactions and duration, select friend groups with interaction times exceeding a set threshold, calculate the matching degree index of shared interests within the group, analyze the overall interest shift trend based on the intensity of each friend group's attention to cultural and creative products, and extract the shared interest matching degree index; The calculation formula for the matching degree index of shared interests within a group is as follows: ; in, An indicator of the degree of matching that represents shared interests within the group, Representing users and friends The number of interactions, Representing users and friends Matching degree in the same interest category, Represents an individual in a friend group The intensity of attention paid to cultural and creative product categories, Represents the normalized adjustment factor of overall user social interaction, Represents the total number of friends who meet the interaction frequency threshold, Represents the total number of people in the user's friend group: Formula used to calculate the matching index of shared interests within a group This indicator is determined by evaluating the degree of interest matching between the user and a group of friends who meet the interaction frequency threshold. The specific calculation method is as follows: (Users and friends Interactions): Data acquisition method: Use the social platform API to capture the interaction records between users and each friend within a specific time period, such as comments, likes, and message exchanges.

[0045] Example: A user interacts with friend A 30 times in one month.

[0046] (Users and friends Matching in the same interest category): Data Acquisition Method: Calculate the similarity between the user and each of their friends in categories of common interest (e.g., cultural products). This can be done by analyzing their browsing and purchasing behavior data, using methods such as cosine similarity.

[0047] Example: The matching degree between the user and friend A in cultural products is 0.75.

[0048] (Individuals in a friend group Intensity of attention on cultural and creative product categories): Data acquisition method: Analyze friends Based on the purchase records of the users, the ratio of their purchase times of cultural and creative products to their total purchase times is calculated.

[0049] Example: Friend A's purchases of cultural and creative products account for 20% of his total purchases, i.e. .

[0050] (Normalization adjustment factor for overall user social interaction): Basis: To balance the impact of different users' social activity levels, the average number of interactions is used as an adjustment factor.

[0051] Example: The average monthly interactions per user is 50, which is used as a normalization adjustment factor .

[0052] and : : The total number of friends in the user group that meet the interaction frequency threshold. Assume it is 5.

[0053] : The total number of people in the user's friend group, assumed to be 10.

[0054] Calculation process example: Assuming that only one friend meets the threshold condition, the calculation process is as follows: ; ; ; Result interpretation: The calculation results This indicates that after considering the frequency of interaction between users and their friends and the degree of interest matching, this user group exhibits a moderate degree of compatibility in terms of shared interests. This suggests that while there is some shared interest, there is still room for improvement. This metric can be used to optimize recommendation algorithms on social platforms to increase user satisfaction and engagement.

[0055] S303: Invoke the shared interest matching index, combine the user's social interaction level, adjust the influence ratio of the recommended product, adjust the recommendation order according to the degree of association between the user and friends at different levels, and obtain the social association recommendation ratio; Based on the user's social interaction level, the influence proportion of recommended products is adjusted, the interaction level is set, and friends are stratified according to the interaction frequency. The stratification method is set as follows: high-frequency interaction friends (interaction frequency ≥ 2 times / day), medium-frequency interaction friends (0.5≤interaction frequency < 2 times / day), and low-frequency interaction friends (interaction frequency < 0.5 times / day). The recommendation order is adjusted according to the degree of association between the user and friends at different levels, and the social association recommendation proportion is calculated. The recommendation proportion calculation method is set as follows: social association recommendation proportion = interaction level weight × interest fit. For example, if the weight of high-frequency interaction friends is set to 0.5, the weight of medium-frequency interaction friends is set to 0.3, and the weight of low-frequency interaction friends is set to 0.2, and the interest fit of a user in the high-frequency friend group is 0.85, then the recommendation proportion of high-frequency interaction friends is calculated as 0.5×0.85=0.425. After calculating the recommendation proportions of friends at all levels, they are normalized so that the total recommendation proportion is 1. Finally, the user's social association recommendation proportion is obtained for the recommendation system to adjust the personalized recommendation strategy.

[0056] The specific steps of S4 are: S401: Obtain the user's geographic location, access time, weather conditions, and device type, call the latitude and longitude data of the user's location, access time, current weather type, and device model, calculate the feature matching degree of the user's scene, and obtain the user scene feature matching value; Access time, weather conditions, device type, call the user's current latitude and longitude information, read the user's city and specific location information, call the access time, record the timestamp and convert it to local time format, extract weather condition data, read the current weather category, such as sunny, rainy, snowy, etc., call the user's device model, extract device category information, such as mobile phone, tablet, PC, etc., calculate the feature matching degree of the user's scene, and set the scene matching calculation method as: scene matching degree = (geographic weight × geographic similarity) + (time weight × time matching degree) + (weather weight × weather adaptability) + (device weight × device adaptability), among which the geographic similarity calculation is the user The distance between the current location and the historically visited locations is normalized. For example, if the user's current coordinates are (30.5, 120.2) and the coordinates of the historically visited points are (30.6, 120.3), the distance between the two points is calculated and normalized. The time matching degree is calculated as the overlap ratio between the user's current visit time and the historically high-frequency visit time. The weather adaptability is calculated as the user's historical click preference ratio under different weather conditions. The device adaptability is calculated as the user's historical browsing time ratio on the device type. For example, if the user's historical browsing time ratio on the PC is 0.65, the device adaptability is 0.65. Finally, the user scenario feature matching value is calculated for subsequent recommendation strategy optimization.

[0057] S402: Invoke the user scenario feature matching value, adjust the recommended categories of cultural and creative products based on the user's location, invoke the category tag of the user's current location, filter cultural and creative products with high category tag correlation, modify the display structure of recommended content based on the visit time period, invoke the product click preference values ​​for different time periods, filter high-preference categories and optimize the display order, filter suitable products based on weather conditions, extract the matching factors between weather conditions and products, adjust the display ratio in the recommendation list, and obtain the scenario adaptation adjustment weight; Adjust the recommended categories of cultural and creative products based on the user's location, call the category label of the user's current geographic location, extract the cultural, commercial, leisure and other scene labels of the user's location, filter out cultural and creative products with high correlation with category labels, and set the category correlation calculation method as follows: Category correlation = historical purchase conversion rate of products in this category at this location. For example, in a certain shopping mall, the conversion rate of cultural products is 0.35, and the conversion rate of leisure products is 0.20, then the category correlation is ranked as cultural > leisure. Modify the display structure of the recommended content based on the visit time period, call the user's historical click behavior data in this time period, filter out high-click categories and optimize the display order, and set the click preference value calculation method as follows: Click preference value = number of clicks of the category in this time period / total number of clicks in this time period. For example, a user A user clicks on cultural products 30 times and other categories of products 70 times during the time period of 18:00-20:00. The click preference value is calculated as 30 / 100=0.3. Categories with higher preference values ​​are filtered, and suitable products are filtered according to weather conditions. The matching factor between weather conditions and products is extracted, and the matching factor calculation method is set as follows: matching factor = the purchase probability of the category under the current weather. For example, on a rainy day, the purchase probability of rain gear products is 0.50, while the purchase probability of decorative products is 0.15. The matching factor ranking is rain gear > decorative category. The display ratio in the recommendation list is adjusted, and the display ratio adjustment method is set as follows: final recommendation ratio = (category correlation × time preference value × weather matching factor) / normalization factor. After calculation, the scene adaptation adjustment weight is obtained for subsequent recommendation ranking optimization.

[0058] S403: Calling the scene adaptation adjustment weight, dynamically adjusting the scene matching weight, combining the user's access scene characteristics and the recommended adjustment coefficient, calculating the adjustment proportion of the environment adaptation recommendation, and obtaining the environment adaptation adjustment coefficient; Dynamically adjust the scene matching weight, and set the dynamic weight adjustment method as follows: Adjustment weight = (scene matching degree × visit frequency coefficient) + (historical purchase ratio × purchase conversion coefficient) + (recommendation click feedback × user interaction coefficient), where the visit frequency coefficient is calculated as the proportion of the user's visits in the scene. For example, the number of visits of the user in the shopping mall scene is 50, and the total number of visits is 200, then the visit frequency coefficient is calculated as 50 / 200=0.25. The purchase conversion coefficient is calculated as the proportion of purchase behavior in the scene. For example, the number of purchases in the scene is 20, and the total number of purchases is 20. is 100, then the purchase conversion coefficient is calculated as 20 / 100=0.2. The recommendation click feedback is calculated as the proportion of recommended product clicks. For example, if the recommended product has 40 clicks and the total number of clicks is 80, then the recommendation click feedback is calculated as 40 / 80=0.5. Combined with the user access scenario characteristics and the recommendation adjustment coefficient, the adjustment proportion of the environment adaptation recommendation is calculated. The adjustment proportion calculation method is set as: environment adaptation adjustment coefficient = (scene adaptation adjustment weight × dynamic adjustment weight) / normalization coefficient. Finally, the environment adaptation adjustment coefficient is obtained for the recommendation system to optimize the weight distribution of recommended content.

[0059] The specific steps of S5 are: S501: Obtain the user interest offset weight, social association recommendation ratio, and environment adaptation adjustment coefficient, call each weight parameter, calculate the ranking weight of the recommended content, calculate the ranking score based on the influence ratio of each weight parameter, and obtain the recommendation ranking weight; Obtain the user interest offset weight, social association recommendation ratio, and environment adaptation adjustment coefficient, call each weight parameter, normalize all weight parameters, ensure that each weight value is in the same calculation range, set the normalization calculation method to: normalized weight = (original weight - minimum weight) / (maximum weight - minimum weight), calculate the ranking weight of the recommended content, call the normalized weight parameter, and calculate the ranking score based on the influence ratio of each weight parameter. Set the ranking score calculation method to: ranking score = (interest offset weight × interest influence ratio) + (social association recommendation ratio × social influence ratio) For example, if a user's interest offset weight is 0.4, the social association recommendation ratio is 0.6, and the environment adaptation adjustment coefficient is 0.5, the interest influence ratio is set to 0.3, the social influence ratio is 0.4, and the scene influence ratio is 0.3, then the recommendation ranking score of the user is calculated as (0.4×0.3)+(0.6×0.4)+(0.5×0.3)=0.12+0.24+0.15=0.51. After calculating the recommendation ranking scores of all users, they are sorted in descending order according to the ranking scores to obtain the recommendation ranking weight.

[0060] S502: Invoke the recommendation ranking weight, filter products with high recent interaction frequency, and increase recommendation priority. Invoke the user's recent browsing, collection, and purchase records to calculate the interaction frequency of each product. Filter high-frequency products based on the interaction frequency threshold. If the attention of each product decreases, call the access trend to calculate the decline rate, adjust the ranking score, calculate the interaction stability based on the user's continuous interaction, modify the recommendation display strategy, and obtain the optimized recommendation ranking result. Filter products with high recent interaction frequency to improve recommendation priority, call the user's recent browsing, collection, and purchase records, calculate the interaction frequency of each product, and set the interaction frequency calculation method as: interaction frequency = (number of views + number of collections + number of purchases) / statistical period. For example, a user browsed a product 10 times, collected it 3 times, and purchased it 2 times in the past 7 days. The interaction frequency of this product is calculated as (10+3+2) / 7=2.14. Filter high-frequency products based on the interaction frequency threshold. Set the threshold to 1.5 times the average interaction frequency of all products. For example, if the average interaction frequency of all products is 1.8, then the threshold is calculated as 1.8×1.5=2.7. Filter products with an interaction frequency ≥2.7. If the attention of a product decreases, call the access trend to calculate the decline. Set the decline calculation method as: decline = (interaction frequency of the previous period - interaction frequency of the current period) For example, the interaction frequency of a product in the previous 7 days is 3.2, and the interaction frequency in the current 7 days is 2.5, then the decline is calculated as (3.2-2.5) / 3.2=0.22. If the decline exceeds the set threshold, the recommendation ranking weight is reduced, and the interaction stability is calculated based on the user's continuous interaction. The interaction stability calculation method is set as: interaction stability = 1-standard deviation of interaction frequency / mean of interaction frequency. For example, the interaction frequency of a user for a certain product in four consecutive cycles is 2.8, 3.0, 2.9, and 2.7, respectively. The standard deviation is calculated to be 0.12, and the mean is 2.85. The interaction stability is calculated as 1-0.12 / 2.85=0.96. If the interaction stability is close to 1, it means that the interaction is relatively stable. If it is lower than the set threshold, the recommendation display strategy is adjusted to finally obtain the optimized recommendation ranking result.

[0061] S503: Call the optimization recommendation sorting results, adjust the order of recommended content according to product priority, calculate the content display weight based on user interaction preferences, optimize the display structure, and obtain a product personalized analysis plan; Adjust the order of recommended content based on product priority, rearrange recommended content according to the optimized ranking score, combine user interaction preferences, call users' historical interaction records on products of different categories, calculate content display weight, and set the display weight calculation method as follows: Display weight = (optimized recommendation ranking score × ranking influence ratio) + (historical interaction preference × interaction influence ratio) + (recommendation click feedback × click influence ratio), where the ranking influence ratio, interaction influence ratio, and click influence ratio are set to 0.4, 0.4, and 0.2, respectively. For example, if a product has an optimized recommendation ranking score of 0.65, a historical interaction preference score of 0.7, and a recommended click feedback score of 0.6, then the display weight is calculated as (0.65×0.4)+(0.7×0.4)+(0.6×0.2)=0.26+0.28+0.12=0.66. After calculating the display weights of all products, optimize the display structure, adjust the order of recommended products based on the display weight, and obtain a product personalized analysis plan.

[0062] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A personalized analysis method for cultural and creative products based on collaborative filtering, characterized by: The following steps are involved: S1: Obtain user browsing, collection, and purchase behavior data, calculate the changes in adjacent behavior intervals, adjust the behavior impact based on browsing frequency, collection density, and purchase conversion, calculate the behavior contribution ratio based on collection frequency and stay time, extract short-term behavior trends, and obtain user behavior change trend parameters; S2: Based on the user behavior change trend parameters, calculate the recent behavior change amplitude, filter out categories with obvious fluctuations, extract key interest transfer points, calculate the matching weight of individual user behavior and group interest, adjust the recommendation ranking, and obtain the user interest shift weight; S3: Obtain user social interaction data, calculate the interest fit between the user and their friends, select friend groups with high interaction and similar interests, calculate the shared interest matching index, adjust the recommendation weight based on the friend group's attention intensity for cultural and creative products, and optimize the recommendation based on the social hierarchy to obtain the proportion of social related recommendations; S4: Obtain the user's geographic location, access time, weather, and device type, calculate the scene matching degree, adjust the recommendation category based on the user's location, modify the recommendation structure based on time, screen suitable products, dynamically adjust the matching weight, and obtain the environment adaptation adjustment coefficient.

2. The method for personalized analysis of cultural and creative products based on collaborative filtering according to claim 1, characterized in that: The user behavior change trend parameters include time interval change value, behavior impact weight, category contribution ratio, and short-term trend index; the user interest shift weight includes behavior change amplitude, interest transfer point, individual matching weight, and recommendation ranking weight; The social association recommendation ratio includes interest fitting degree, shared interest matching degree, friend attention intensity, and social interaction level weight; the environmental adaptation adjustment coefficient includes scene matching degree, recommendation category adjustment coefficient, display structure correction parameter, and product screening weight.

3. The method for personalized analysis of cultural and creative products based on collaborative filtering according to claim 1, characterized in that: The specific steps to obtain user browsing, collection, and purchase behavior data, calculate the changes in adjacent behavior intervals, adjust the behavior impact based on browsing frequency, collection density, and purchase conversion, calculate the behavior contribution ratio based on collection frequency and stay time, extract short-term behavior trends, and obtain user behavior change trend parameters are as follows: S101: Obtain the user's browsing, collection, and purchase behavior data, arrange them in chronological order, calculate the time intervals between adjacent behaviors, call the time points of the previous and next behaviors to calculate the time difference, use the time interval mean of the behavior sequence as a benchmark, determine the change of each time interval, and generate the behavior time interval change coefficient; S102: Calculate the influence of the behavior category on the overall behavior sequence by calling the behavior time interval variation coefficient and combining it with browsing frequency, collection density, and purchase conversion. Increase the influence when the interval is shortened and decrease the influence when the interval is increased. Calculate the contribution ratio based on the occurrence ratio of the behavior category and the time interval variation trend to generate the behavior category influence ratio value. S103: Call the behavior category influence weight value, combine it with the collection frequency and page dwell time, analyze the short-term changes of the behavior category, call the behavior category influence weight value of adjacent time periods, calculate the increase and decrease rate, extract the short-term behavior change trend, and generate user behavior change trend parameters.

4. The method for personalized analysis of cultural and creative products based on collaborative filtering according to claim 1, characterized in that: Based on the user behavior change trend parameters, the recent behavior change amplitude is calculated, the categories with obvious fluctuations are screened, key interest transfer points are extracted, the matching weights of individual user behavior and group interest are calculated, and the recommendation ranking is adjusted to obtain the user interest shift weights. S201: Obtain the user behavior change trend parameter, extract the recent behavior change rate, set the behavior time interval, calculate the change range of each category of behavior, call the user's behavior record data, calculate the increase or decrease range based on the change amount of each category of behavior within the time interval, and obtain the behavior category change range; S202: Calculating the change range of the behavior categories, screening the behavior categories with obvious fluctuations based on the number of favorites, browsing paths, and purchase times, calculating the fluctuation range of each category of behavior within a set time interval, calculating the category change ratio, extracting the behavior categories whose change rate exceeds the set threshold, identifying the critical time points of the behavior category changes, and extracting the key interest transfer points; S203: Call the key interest transfer point, calculate the matching weight of individual user behavior and group interest, adjust the priority of recommended content based on the changing trend of user behavior category and analyze the degree of deviation from group interest, and obtain the user interest deviation weight.

5. The method for personalized analysis of cultural and creative products based on collaborative filtering according to claim 4, characterized in that: The specific calculation formula for the matching weight between individual user behavior and group interests is: ； in, Represents the matching weight between individual user behavior and group interests, Represents the user in the category The frequency of behaviors on Representative groups in categories The average frequency of behaviors on Represents the user in the category The weight of the behavioral change trend, Represents the user in the time period The behavioral deviation value within Represents the normalized adjustment factor for matching all categories of user behavior. Represents the total number of user behavior categories, Represents the total number of user behavior time periods.

6. The method for personalized analysis of cultural and creative products based on collaborative filtering according to claim 1, characterized in that: Obtain user social interaction data, calculate the interest fit between the user and friends, screen friend groups with high interaction and similar interests, calculate shared interest matching indicators, adjust recommendation weights based on the intensity of friend groups' attention to cultural and creative products, and optimize recommendations based on social hierarchies. The specific steps to obtain the proportion of social related recommendations are as follows: S301: Obtain the user's social interaction data, extract the browsing, collection, and purchase records of friends, call the user and friends' behavior data, calculate the matching degree based on the number of browsing, collection, and purchase behaviors of the same category, and obtain the interest fit between the user and friends; S302: Calling the interest fit between the user and their friends, screening friend groups with high interaction frequency and similar interests, calling the interaction records between the user and each friend, recording the number of interactions and duration, screening friend groups with interaction times exceeding a set threshold, calculating the matching degree index of shared interests within the group, analyzing the overall interest shift trend based on the intensity of each friend group's attention to cultural and creative products, and extracting the shared interest matching degree index; S303: Calling the shared interest matching degree index, combining the user's social interaction level, adjusting the influence proportion of the recommended product, adjusting the recommendation order according to the degree of association between the user and friends of different levels, and obtaining the social association recommendation proportion.

7. The method for personalized analysis of cultural and creative products based on collaborative filtering according to claim 6, characterized in that: The specific calculation formula for the matching degree index of shared interests within the group is: ； in, An indicator of the degree of matching that represents shared interests within the group, Representing users and friends The number of interactions, Representing users and friends Matching degree in the same interest category, Represents an individual in a friend group The intensity of attention paid to cultural and creative product categories, Represents the normalized adjustment factor of overall user social interaction, Represents the total number of friends who meet the interaction frequency threshold, Represents the total number of people in the user's friend group.

8. The method for personalized analysis of cultural and creative products based on collaborative filtering according to claim 1, characterized in that: The specific steps to obtain the user's geographic location, access time, weather, and device type, calculate the scene matching degree, adjust the recommended categories based on the user's location, modify the recommendation structure based on time, screen suitable products, and dynamically adjust the matching weight to obtain the environment adaptation adjustment coefficient are as follows: S401: Obtain the user's geographic location, access time, weather conditions, and device type, call the latitude and longitude data of the user's location, access time, current weather type, and device model, calculate the feature matching degree of the user's scene, and obtain the user scene feature matching value; S402: Utilize the user scenario feature matching value, adjust the recommended categories of cultural and creative products based on the user's location, utilize the category tag of the user's current location, filter cultural and creative products with high correlation with the category tag, modify the display structure of recommended content based on the visit time period, utilize the product click preference values ​​for differentiated time periods, filter high-preference categories and optimize the display order, filter suitable products based on weather conditions, extract weather condition and product matching factors, adjust the display ratio in the recommendation list, and obtain scenario adaptation adjustment weights; S403: Call the scene adaptation adjustment weight, dynamically adjust the scene matching weight, combine the user access scene characteristics and the recommendation adjustment coefficient, calculate the adjustment proportion of the environment adaptation recommendation, and obtain the environment adaptation adjustment coefficient.

9. The method for personalized analysis of cultural and creative products based on collaborative filtering according to claim 1, characterized in that: The method also Including, S5: based on the user interest shift weight, the social association recommendation ratio and the environment adaptation adjustment coefficient, calculating the recommendation ranking weight, screening products with high recent interaction frequency and increasing their priority, lowering the ranking if the attention of similar products decreases, adjusting the display strategy in combination with continuous interaction, optimizing the arrangement order, and obtaining a product personalized analysis plan; The product personalization analysis solution includes recommendation ranking weight, interaction frequency coefficient, attention adjustment parameter, and display optimization strategy.

10. The method for personalized analysis of cultural and creative products based on collaborative filtering according to claim 9, characterized in that: Based on the user interest shift weight, the social association recommendation ratio and the environment adaptation adjustment coefficient, the recommendation ranking weight is calculated, and products with high recent interaction frequency are screened for priority. If the attention of similar products decreases, the ranking is lowered. Combined with continuous interaction, the display strategy is adjusted and the arrangement order is optimized. The specific steps of obtaining a product personalized analysis solution are as follows: S501: Obtain the user interest offset weight, the social association recommendation ratio, and the environment adaptation adjustment coefficient, call each weight parameter, calculate the ranking weight of the recommended content, calculate the ranking score based on the influence ratio of each weight parameter, and obtain the recommendation ranking weight; S502: Utilize the recommendation ranking weights to filter products with high recent interaction frequencies and increase their recommendation priority. Utilize the user's recent browsing, collection, and purchase records to calculate the interaction frequency of each product. Filter high-frequency products based on the interaction frequency threshold. If the attention of each product decreases, utilize the access trend to calculate the decrease rate and adjust the ranking score. Calculate the interaction stability based on the user's continuous interaction status, modify the recommendation display strategy, and obtain an optimized recommendation ranking result. S503: Call the optimization recommendation ranking result, adjust the order of recommended content according to product priority, calculate content display weight based on user interaction preferences, optimize the display structure, and obtain a product personalized analysis plan.

Images