Method and system for personalized recommendation of cultural and creative products based on user portraits

Abstract

The invention relates to the technical field of intelligent recommendation, and provides a personalized recommendation method and system for cultural and creative products based on a user portrait, and the method comprises the steps: carrying out the stay duration marking of coordinate points in browsing track data, and obtaining a visual focusing region; performing pixel color analysis and peak detection on the visual focusing area to obtain a cultural image color label; performing word frequency statistics on the historical order data to obtain dominant category preference tags; semantic alignment and intention analysis are carried out on the culture image color labels and the dominant category preference labels to obtain composite culture preferences; based on the composite culture preference, preliminarily screening and sorting the to-be-recommended cultural and creative products to obtain preliminarily recommended products; performing product image segmentation on the preliminarily recommended product to obtain a pattern shape code of the preliminarily recommended product, and performing collaborative screening matching on the pattern shape code and the cultural image color label to obtain a personalized recommended product; according to the method, the personalized recommendation efficiency of the cultural and creative products can be improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent recommendation technology, and in particular to a method and system for personalized recommendation of cultural and creative products based on user profiles. Background Technology

[0002] In the field of cultural and creative product recommendations, existing technologies only delve into superficial category preferences when exploring user needs. They fail to integrate user visual browsing characteristics to uncover underlying cultural imagery preferences, thus hindering the construction of a multi-dimensional system of user cultural preferences. Consequently, recommendations only match users' explicit consumption needs, failing to resonate with their deeper cultural aesthetics and personalized demands. Furthermore, existing technologies lack the collaborative screening of product visual features and user preferences during the product matching process, relying solely on simple category matching to complete recommendations. This results in a low degree of alignment between recommended cultural and creative products and users' actual preferences, failing to achieve truly personalized recommendations.

[0003] Current technologies for recommending cultural and creative products lack standardized feature extraction and fusion processes for user data. Browsing trajectory data lacks refined visual focus analysis, and word frequency statistics of historical order data do not consider the weight decay over time, resulting in insufficient accuracy in user profile construction. Furthermore, the matching calculation method for product recommendations is relatively simplistic, lacking a multi-dimensional collaborative matching model based on patterns, colors, and categories, leading to low recommendation efficiency and accuracy. Therefore, improving the accuracy and efficiency of personalized recommendations for cultural and creative products has become an urgent problem to be solved. Summary of the Invention

[0004] This invention provides a method and system for personalized recommendation of cultural and creative products based on user profiles, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, this invention provides a personalized recommendation method for cultural and creative products based on user profiles, comprising: Step a: Collect the user's browsing trajectory data and mark the dwell time of the coordinate points in the browsing trajectory data to obtain the user's visual focus area; Step b: Perform pixel color analysis on the visual focus area to obtain the user's color distribution histogram, and perform peak detection on the color distribution histogram to obtain the user's cultural image color tag; Step c: Perform word frequency statistics on the user's historical order data to obtain the user's explicit category preference tags; Step d: Semantically align the cultural image color tags with the explicit category preference tags to obtain the user's preference fusion features, and perform intent analysis on the preference fusion features to obtain the user's composite cultural preferences; Step e: Based on the aforementioned composite cultural preferences, perform an initial screening and ranking of the user's recommended cultural and creative products to obtain the user's preliminary recommended products; Step f: Perform product image segmentation on the initially recommended products to obtain the pattern shape code of the initially recommended products, and perform collaborative screening and matching of the pattern shape code with the cultural image color tag to obtain the personalized recommended products for the user.

[0006] In a preferred embodiment, the step of collecting the user's browsing trajectory data and marking the dwell time of coordinate points in the browsing trajectory data to obtain the user's visual focus area includes: Obtain the user's browsing interface coordinate stream data, and remove noise points from the browsing interface coordinate stream data to obtain the user's valid browsing trajectory point sequence; The time stamps of the trajectory points in the valid browsing trajectory point sequence are parsed to obtain the spatiotemporal dwell segment of the user; The coordinates of the first and last trajectory points of the spatiotemporal dwell segment are merged to obtain the gaze center point of the spatiotemporal dwell segment; The duration of the gaze center point is determined based on the number of trajectory points within the spatiotemporal dwell segment, the first trajectory point, and the last trajectory point. The duration of the gaze is assigned as attribute information to the gaze center point to obtain the user's visual focus area.

[0007] In a preferred embodiment, the step of performing pixel color analysis on the visual focus area to obtain the user's color distribution histogram, and performing peak detection on the color distribution histogram to obtain the user's cultural imagery color tag, includes: Extract the gaze center point in the visual focus area, and define a color sampling window based on the gaze center point; Collect the pixel color channel values ​​in the color sampling window, and deduplicate and merge the pixel color channel values ​​to obtain the user's gaze point color features; The color features of the gaze point are frequency-collected to obtain the color frequency distribution histogram of the user. Peak values ​​are extracted from the frequency distribution histogram of the color system to obtain the user's significant color system categories; The significant color categories are associated and matched with a preset cultural imagery mapping library to obtain the user's cultural imagery color tags.

[0008] In a preferred embodiment, the step of performing word frequency statistics on the user's historical order data to obtain the user's explicit category preference tags includes: Extract and parse the user's historical order data to obtain the product title text of the historical order data; Stop words are removed from the product title text to obtain the user's product keyword sequence; The product keyword sequence is matched and mapped with a preset cultural and creative product category dictionary to obtain the user's effective category keywords; The frequency of the effective category keywords in the product keyword sequence is counted to obtain the original frequency of the effective category keywords; Based on the original frequency, a time decay coefficient is assigned to the effective category keywords to obtain the weighted time-effect frequency of the effective category keywords; Based on the weighted timeliness frequency, the effective category keywords are sorted and filtered to obtain the user's high-frequency core category keywords; The high-frequency core category words are associated and aligned with the category hierarchy structure in the cultural and creative category dictionary to obtain the user's explicit category preference tags.

[0009] In a preferred embodiment, the semantic alignment of the cultural imagery color tags with the explicit category preference tags to obtain the user's preference fusion features includes: Semantic elements are extracted from the cultural imagery color tags and the explicit category preference tags to obtain the user's color semantic features and category semantic features; Based on the color semantic features, the category semantic features are matched with the same domain semantic features to obtain the semantic matching representation of the user; Based on the semantic matching representation, the color semantic features and the category semantic features are fused and encoded to obtain the user's preference fusion features.

[0010] In a preferred embodiment, the intention analysis performed on the preference fusion features to obtain the user's composite cultural preferences includes: By performing cultural and creative semantic intent mining on the aforementioned preference fusion features, the cultural intention representation of the user can be obtained. Based on the cultural intention representation, the cultural attribute association of the user's cultural and creative needs is determined to obtain the user's cultural attribute association representation. Based on the cultural attribute association representation, the user's cultural and creative preferences are constructed in an integrated manner to obtain the user's composite cultural preferences.

[0011] In a preferred embodiment, the initial screening and ranking of the user's recommended cultural and creative products based on the composite cultural preferences to obtain the user's preliminary recommended products includes: Based on the composite cultural preference profile, the cultural attribute matching search is performed on the cultural and creative products to be recommended to the user to obtain the matching degree of the user's cultural and creative products. Based on the matching degree of the cultural and creative products, the cultural and creative products to be recommended are initially screened for suitability to obtain the user's initial screened cultural and creative products. The cultural and creative products initially screened are ranked according to their cultural preference compatibility to obtain the user's preliminary recommended products.

[0012] In a preferred embodiment, the step of performing product image segmentation on the initially recommended products to obtain the pattern shape code of the initially recommended products, and then performing collaborative screening and matching of the pattern shape code with the cultural imagery color tag to obtain the personalized recommended products for the user, includes: Perform foreground-background separation processing on the product image of the initially recommended product to obtain the main pattern image of the initially recommended product; The outline features of the pattern main image are extracted to obtain the pattern outline features of the preliminary recommended product; The pattern outline features are structured and encoded to obtain the pattern shape code of the preliminary recommended product; Calculate the degree of coordination between the pattern shape code and the cultural image color tag; Based on the collaborative matching degree, the initially recommended products are collaboratively screened to obtain personalized recommended products for the user.

[0013] In a preferred embodiment, the formula for calculating the collaborative matching degree is as follows: ; in, The degree of collaborative matching, The feature value encoded by the pattern shape, The feature value of the cultural image color tag, The category weight of the explicit category preference label. The cultural preference weights for the aforementioned cultural image color labels. This represents the domain correlation coefficient between pattern and color.

[0014] To address the aforementioned problems, the present invention also provides a personalized recommendation system for cultural and creative products based on user profiles, the system comprising: The visual focus area extraction module is used to collect the user's browsing trajectory data and mark the dwell time of the coordinate points in the browsing trajectory data to obtain the user's visual focus area. The cultural imagery color tag generation module is used to perform pixel color analysis on the visual focus area to obtain the user's color distribution histogram, and to perform peak detection on the color distribution histogram to obtain the user's cultural imagery color tag. The explicit category preference tag determination module is used to perform word frequency statistics on the user's historical order data to obtain the user's explicit category preference tags; The composite cultural preference construction module is used to semantically align the cultural image color tags with the explicit category preference tags to obtain the user's preference fusion features, and to perform intent analysis on the preference fusion features to obtain the user's composite cultural preferences. The preliminary recommended product screening module is used to perform preliminary screening and sorting of the user's cultural and creative products based on the composite cultural preferences, so as to obtain the user's preliminary recommended products; The personalized product recommendation module is used to perform product image segmentation on the initially recommended products to obtain the pattern shape code of the initially recommended products, and to perform collaborative screening and matching of the pattern shape code with the cultural image color tag to obtain the personalized recommended products for the user.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention significantly improves recommendation accuracy by collecting and analyzing user data from multiple dimensions. From extracting visual focus areas from browsing trajectories to generating cultural imagery color tags through color analysis, and then combining explicit category preference tags from historical order data, the semantic alignment and intent analysis of the two tags are completed. This constructs a composite cultural preference that fits user needs, allowing the initial screening and ranking of cultural and creative products to be accurately guided by user needs. At the same time, by obtaining pattern shape codes through product image segmentation and co-screening and matching them with cultural imagery color tags, a deep fit between products and user preferences is achieved from multiple dimensions of color, category, and pattern, greatly improving the matching accuracy of personalized recommendations.

[0016] 2. This invention achieves its goals through standardized data analysis and feature processing. From noise removal from browsing data and weight allocation of time decay coefficients to formulaic calculation of collaborative matching degree, the entire recommendation process is freed from the interference of subjective judgment, realizing data-driven intelligent recommendation. At the same time, each functional module performs its own function while cooperating with each other, forming a complete technical link from user feature extraction to final product recommendation. This effectively improves the overall execution efficiency of personalized recommendation of cultural and creative products and makes the recommendation results more objective and reasonable. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a personalized recommendation method for cultural and creative products based on user profiles, provided in an embodiment of the present invention. Figure 2 A functional module diagram of a personalized recommendation system for cultural and creative products based on user profiles, provided in an embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0019] This application provides a method for personalized recommendation of cultural and creative products based on user profiles. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for personalized recommendation of cultural and creative products based on user profiles can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0020] Reference Figure 1 The diagram shown is a flowchart illustrating a personalized recommendation method for cultural and creative products based on user profiles, according to an embodiment of the present invention. In this embodiment, the personalized recommendation method for cultural and creative products based on user profiles includes: Step a: Collect the user's browsing trajectory data and mark the dwell time of the coordinate points in the browsing trajectory data to obtain the user's visual focus area; In this embodiment of the invention, the step of collecting the user's browsing trajectory data and marking the dwell time of the coordinate points in the browsing trajectory data to obtain the user's visual focus area includes: Obtain the user's browsing interface coordinate stream data, and remove noise points from the browsing interface coordinate stream data to obtain the user's valid browsing trajectory point sequence; The time stamps of the trajectory points in the valid browsing trajectory point sequence are parsed to obtain the spatiotemporal dwell segment of the user; The coordinates of the first and last trajectory points of the spatiotemporal dwell segment are merged to obtain the gaze center point of the spatiotemporal dwell segment; The duration of the gaze center point is determined based on the number of trajectory points within the spatiotemporal dwell segment, the first trajectory point, and the last trajectory point. The duration of the gaze is assigned as attribute information to the gaze center point to obtain the user's visual focus area.

[0021] Obtain the user's browsing interface coordinate stream data, and remove noise points from the browsing interface coordinate stream data to obtain the user's valid browsing trajectory point sequence.

[0022] The system continuously captures coordinate data generated by every user action within the browsing interface, fully recording the timestamp information corresponding to each coordinate point. This forms a browsing interface coordinate stream data arranged in chronological order. A data index is simultaneously built during data acquisition for easy subsequent retrieval and filtering. Abnormal data in the coordinate stream is identified and processed. Specifically, coordinate points exceeding the effective display area of ​​the browsing interface are filtered out. These coordinate points reflect user accidental touches outside the interface or actions without actual browsing significance, and are judged as invalid noise data. Coordinate points generated repeatedly under the same timestamp are also identified; these are redundant information collected repeatedly and are also classified as noise points to be removed. All identified invalid noise data is completely removed from the original coordinate stream. The remaining coordinate points are the true trajectory points reflecting the user's effective browsing behavior. These valid trajectory points are then reordered and organized according to the original time sequence of acquisition, ultimately forming a standardized and non-redundant sequence of valid user browsing trajectory points. Each trajectory point in this sequence has a unique timestamp and accurate interface coordinates, accurately reconstructing the user's actual movement path within the interface.

[0023] The system reads the timestamp information of each trajectory point in the valid browsing trajectory point sequence one by one. It calculates the difference between the timestamps of two adjacent trajectory points and determines whether the time difference is within a pre-set reasonable time threshold range. If the time difference is within the threshold range, the two adjacent trajectory points are considered to be continuous valid browsing behavior trajectories. Such continuous trajectory points are merged and integrated to form an independent browsing segment unit. If the time difference exceeds the threshold range, it is determined that there is an interruption in the browsing behavior between the two trajectory points. Using this as the dividing point, the valid browsing trajectory point sequence is divided into multiple independent segments. Each segment contains several continuous trajectory points with time intervals that meet the requirements. Each independent segment is a spatiotemporal dwell segment. Each spatiotemporal dwell segment covers all the coordinate information of the user's browsing trajectory within that time period and the corresponding time sequence information, which can completely reflect the user's continuous browsing path and time distribution within a certain time period.

[0024] The horizontal and vertical coordinates of the first trajectory point in each spatiotemporal dwell segment are extracted, as are the horizontal and vertical coordinates of the last trajectory point in the segment. The horizontal coordinates of the first and last trajectory points are calculated, and the sum of the two values ​​is divided by two to obtain the horizontal center coordinate of the spatiotemporal dwell segment. The same calculation logic is applied to the vertical coordinates of the first and last trajectory points, and the sum of the two values ​​is divided by two to obtain the vertical center coordinate of the spatiotemporal dwell segment. The calculated horizontal and vertical center coordinates are combined to form a new coordinate point, which is the gaze center point of the spatiotemporal dwell segment. This gaze center point is calculated based on the beginning and end positions of the user's browsing trajectory within the segment, accurately representing the core focus position of the user's visual attention during that time period. This effectively simplifies the complex coordinate data of the user's browsing path and extracts the key visual focus points.

[0025] The system obtains the specific timestamp values ​​corresponding to the first and last trajectory points in the spatiotemporal fixation segment. The timestamp value of the last trajectory point is subtracted from the timestamp value of the first trajectory point; the resulting time difference is the fixation duration corresponding to that fixation center point. The calculation process strictly adheres to the time unit conversion rules to ensure the accuracy and consistency of the time difference. This fixation duration calculation is based solely on the timestamp difference between the first and last trajectory points of the spatiotemporal fixation segment and is not directly related to the number of trajectory points contained within the segment. The number of trajectory points serves only as a reference for verifying the continuity of the segment and does not participate in the specific calculation of the fixation duration. The final determined fixation duration directly reflects the length of time the user's visual fixation is at the fixation center point.

[0026] A correlation is established between the gaze center point and its corresponding dwell time. The calculated dwell time value is bound to the corresponding gaze center point as a unique attribute, so that each gaze center point has a unique coordinate identifier and a corresponding dwell time attribute. The dwell time attribute can intuitively reflect the intensity of the user's visual attention at the focal point. The larger the duration value, the higher the user's attention at that position. All gaze centers with completed attribute binding are uniformly summarized and organized, and arranged according to the user's browsing time order to form a complete set of gaze centers. This set is the user's visual focus area. Each gaze center point in the set corresponds to the core point of the user's visual attention. The dwell time attribute of each point constitutes the key data foundation reflecting the user's visual preferences and attention habits.

[0027] The beneficial effects are as follows: through the refined processing of browsing trajectory data throughout the entire process, from noise removal from the raw data to ensure data authenticity, to the precise capture of continuous user browsing behavior by dividing spatiotemporal dwell segments, to the merging and extraction of core visual points by the center of gaze, and the assignment of precise attention attributes to the points by calculating the dwell time, the user's visual focus area is accurately constructed. This construction process does not rely on complex auxiliary tools and can be completed through standardized coordinate processing and time calculation, ensuring the objectivity and accuracy of visual focus area extraction. This provides reliable core data support for subsequent analysis of user cultural imagery and color preferences, effectively improving the efficiency and quality of user visual feature extraction, avoiding subjective errors caused by manual judgment, and ensuring the rigor and stability of the entire user preference analysis process.

[0028] Step b: Perform pixel color analysis on the visual focus area to obtain the user's color distribution histogram, and perform peak detection on the color distribution histogram to obtain the user's cultural image color tag; In this embodiment of the invention, the step of performing pixel color analysis on the visual focus area to obtain the user's color distribution histogram, and performing peak detection on the color distribution histogram to obtain the user's cultural imagery color tag, includes: Extract the gaze center point in the visual focus area, and define a color sampling window based on the gaze center point; Collect the pixel color channel values ​​in the color sampling window, and deduplicate and merge the pixel color channel values ​​to obtain the user's gaze point color features; The color features of the gaze point are frequency-collected to obtain the color frequency distribution histogram of the user. Peak values ​​are extracted from the frequency distribution histogram of the color system to obtain the user's significant color system categories; The significant color categories are associated and matched with a preset cultural imagery mapping library to obtain the user's cultural imagery color tags.

[0029] Extract the fixation center point in the visual focus area, and define a color sampling window based on the fixation center point.

[0030] The system first retrieves the set of visual focus areas and extracts the coordinate data of each gaze center point with a unique dwell time attribute. For each gaze center point, its specific vertical and horizontal positions in the browsing interface are determined. Using this as the core center point, a square color sampling window is defined by expanding outwards according to a preset fixed size range. The boundary of this window is symmetrically distributed around the gaze center point to ensure that the window can completely cover the visual color area around the gaze point. The size of the window is fixed and uniform, ensuring the consistency of the subsequent color sampling range and avoiding color analysis errors caused by different sampling area sizes. Each gaze center point corresponds to an independent color sampling window, and the position of the window is dynamically adjusted according to the coordinates of the gaze center point to accurately cover the core area of ​​the user's visual focus.

[0031] The system traverses the internal area of ​​each color sampling window, reading the red, green, and blue channel values ​​of each pixel in the window row by row and column by column, completely recording the color channel data corresponding to each pixel. The system then summarizes all the collected pixel color channel values ​​to form a raw dataset containing all color information within the window. Based on this, the system identifies and merges the color channel values ​​that appear repeatedly in the dataset, retaining each unique color channel combination and removing duplicate and redundant data. Finally, a set of color channel values ​​without duplicate elements is obtained. This set represents the user's gaze point color feature for that gaze point. Each element in the set represents an independent color combination that has actually appeared in the gaze area, comprehensively and without redundancy reflecting the color composition of the user's gaze area.

[0032] Based on the extracted set of color features of the gaze point, the system counts the total number of times each unique color channel combination in the set actually appears in the pixel data of the original color sampling window. The number of times each color combination appears is its frequency. According to the classification rules of color channel values, combinations with similar color features are classified and integrated to form different color system categories. The system counts all color combinations and their total frequencies included in each color system category, establishing a correspondence between color system and corresponding frequency. Based on the color system category and its corresponding frequency data, a bar chart is constructed in a coordinate system, with the color system category as the horizontal axis and the frequency of the corresponding color system as the vertical axis, to intuitively present the distribution of each color system in the user's visual focus area. This bar chart is the user's color system frequency distribution histogram, which can clearly show the proportion and frequency of different color systems in the user's gaze area.

[0033] The system iterates through each bar in the color frequency distribution histogram, reading the frequency value of each bar. By comparing the frequencies of adjacent bars, it identifies local maxima whose frequencies are significantly higher than those of their left and right neighbors. These local maxima are the peaks in the histogram. All identified peaks are sorted from highest to lowest frequency, and the peak with the highest frequency is selected as the core peak. The color category corresponding to the core peak represents the most frequent and prominent color type in the user's visual focus area. The color categories corresponding to these core peaks are extracted and summarized to form the user's significant color category set. This set accurately identifies the most representative color expression in the user's visual preferences.

[0034] The system pre-constructs a mapping database containing multiple color categories and corresponding cultural imagery terms. Each color category in the database is associated with specific cultural imagery tags, such as elegant, passionate, tranquil, and vibrant, clearly establishing the correspondence between color and cultural connotation. It reads the extracted set of significant color categories and precisely compares each category with the color entries in the mapping database to find perfectly matching color records. It then reads the pre-defined cultural imagery terms under each matching entry and uses these successfully matched terms as the user's cultural imagery color tags. For cases with multiple significant color categories, the system sequentially completes the matching and association for each category. Finally, it integrates all the matched cultural imagery color tags to form a complete user cultural imagery color tag set. This tag set intuitively reflects the user's cultural color preferences and aesthetic inclinations formed during visual browsing.

[0035] The beneficial effects are as follows: By precisely defining the color sampling window, the targeted and consistent area of ​​color collection is ensured. After pixel-level collection, deduplication and merging, and frequency aggregation, a color frequency distribution histogram that accurately reflects the user's visual focus is constructed. Combined with peak detection, the most significant color categories that users are most concerned about are effectively extracted. Finally, through intelligent association with a preset cultural imagery mapping library, the accurate conversion from objective color data to subjective cultural imagery tags is achieved. The entire process is logically rigorous, with each step closely linked, avoiding deviations caused by subjective assumptions. It can accurately capture users' cultural aesthetic preferences based on color, providing solid and reliable color feature support for building a personalized recommendation system that meets users' deep needs. It significantly improves the accuracy and objectivity of user cultural imagery color tag generation, ensuring the quality and efficiency of user profile construction at the front end of the entire recommendation system.

[0036] Step c: Perform word frequency statistics on the user's historical order data to obtain the user's explicit category preference tags; In this embodiment of the invention, the step of performing word frequency statistics on the user's historical order data to obtain the user's explicit category preference tags includes: Extract and parse the user's historical order data to obtain the product title text of the historical order data; Stop words are removed from the product title text to obtain the user's product keyword sequence; The product keyword sequence is matched and mapped with a preset cultural and creative product category dictionary to obtain the user's effective category keywords; The frequency of the effective category keywords in the product keyword sequence is counted to obtain the original frequency of the effective category keywords; Based on the original frequency, a time decay coefficient is assigned to the effective category keywords to obtain the weighted time-effect frequency of the effective category keywords; Based on the weighted timeliness frequency, the effective category keywords are sorted and filtered to obtain the user's high-frequency core category keywords; The high-frequency core category words are associated and aligned with the category hierarchy structure in the cultural and creative category dictionary to obtain the user's explicit category preference tags.

[0037] The system retrieves all historical order data for the user from the user order database. Each historical order is structured and parsed to extract the core text information related to the products. The focus is on obtaining the complete product title text for each product, while retaining the time association information between each product title text and the corresponding order. All extracted product title texts are then compiled and organized to form a text set containing the titles of all cultural and creative products purchased by the user, ensuring that the extracted product title texts are complete and accurate, and accurately correspond to the user's actual consumption information.

[0038] First, a dedicated stop word library for the cultural and creative industry is constructed. This library contains function words, adverbs, conjunctions, and general modifiers in the titles of cultural and creative products that do not have actual category-specific meaning. Then, each product title text after aggregation is broken down word by word. The words after decomposition are compared one by one with the words in the stop word library for the cultural and creative industry. All matching stop words are removed from the title vocabulary, and only effective words with actual cultural and creative category characteristics and reflecting product attributes are retained. The remaining effective words after removing stop words from all product title texts are arranged in the word order of the original title. Then, all the effective words in the titles are integrated to form a continuous word sequence. This sequence is the user's product keyword sequence. The words in the sequence are all core words that can reflect the category characteristics of the products consumed by the user.

[0039] A standardized cultural and creative product category dictionary is pre-built, which includes the category names, product names, and core vocabulary related to all mainstream cultural and creative products. The vocabulary in the dictionary is then categorized. Each word in the user's product keyword sequence is then traversed and precisely matched with all words in the cultural and creative product category dictionary. Keywords that completely match the dictionary words are selected, while non-category-related words that cannot match the dictionary are removed. All successfully matched keywords are summarized, and duplicate words are removed to form a non-redundant vocabulary set. This set is the user's effective category keywords, ensuring that each keyword accurately corresponds to a specific category in the cultural and creative product category dictionary.

[0040] Based on the user's product keyword sequence, each valid category keyword is searched one by one, and the actual number of times each valid category keyword appears in the entire product keyword sequence is recorded. This number is the original frequency of the corresponding valid category keyword. During the statistical process, a full search is performed according to the original word order of the product keyword sequence to ensure that every valid category keyword is accurately counted without omissions or errors. After counting all valid category keywords, a one-to-one correspondence between valid category keywords and their corresponding original frequencies is established, forming a complete original frequency statistical result.

[0041] First, standardized time decay rules are established, dividing the time decay into different levels based on the time interval between the generation time of users' historical orders and the current time. A fixed decay coefficient is set for each decay level, with a smaller decay coefficient for longer time intervals. Then, the generation time of the product orders corresponding to each effective category keyword is retrieved, and the specific time interval between that time and the current time is calculated. According to the time decay rules, a corresponding time decay coefficient is matched for each effective category keyword. The original frequency of the effective category keyword is multiplied by the corresponding time decay coefficient, and the calculated value is the weighted time frequency of the effective category keyword. This calculation is used to distinguish the weight of category keywords for different time consumption behaviors, aligning with the latest consumption preferences of users.

[0042] All valid category keywords are sorted in descending order according to their corresponding weighted time-of-use frequency, with the valid category keyword with the highest weighted time-of-use frequency value placed first. The remaining keywords are then arranged in order of frequency value. A fixed frequency filtering threshold is set, and all valid category keywords with a weighted time-of-use frequency value higher than the threshold are filtered out. If the number of filtered keywords exceeds a preset range, the preset number of keywords at the top of the ranking are selected. The final filtered valid category keywords are then summarized to form the user's high-frequency core category words, ensuring that the category words in this set are all category keywords with high weight and the most representative category keywords in the user's consumption behavior.

[0043] The system retrieves the complete category hierarchy structure from a pre-defined cultural and creative product category dictionary. This structure includes multiple levels of category division, such as major, intermediate, and minor categories of cultural and creative products. Each category term corresponds to its specific position and superior category within the hierarchy. Then, each of the user's high-frequency core category terms is precisely associated with the category hierarchy structure of the cultural and creative product category dictionary. This determines the specific category level and category system of each high-frequency core category term within the hierarchy structure. Based on the category hierarchy structure, the high-frequency core category terms are standardized and tagged. The tagged high-frequency core category terms are then organized according to category hierarchy, and each term is assigned a corresponding category hierarchy attribute. The standardized category terms are used as core tags, and combined with their category hierarchy attributes, a complete tag set is formed. This set represents the user's explicit category preference tags, which accurately and hierarchically reflect the user's consumption preferences for cultural and creative product categories.

[0044] The beneficial effects are as follows: Through refined text processing of user historical order data, from extracting product titles to removing stop words, the purity and effectiveness of category-related keywords are ensured. Combined with the matching mapping of a pre-set cultural and creative category dictionary, standardized screening of category keywords is achieved. Then, through raw frequency statistics and time decay coefficient allocation, a weighted timeliness frequency that fits the user's consumption timeliness is obtained, completing the quantitative analysis of user consumption behavior. Subsequent sorting and screening are aligned with the category hierarchy structure, so that the final generated explicit category preference tags not only fit the characteristics of users' high-frequency consumption of cultural and creative categories, but also have clear category hierarchy attributes. The entire process achieves accurate transformation from users' raw consumption data to standardized category preference tags. The tags can truly and hierarchically reflect users' explicit cultural and creative category consumption preferences, providing accurate and reliable category feature data support for the subsequent construction of user preference fusion features. At the same time, the introduction of the time decay coefficient makes the tags more in line with users' latest consumption preferences, improving the timeliness and accuracy of category preference tags.

[0045] Step d: Semantically align the cultural image color tags with the explicit category preference tags to obtain the user's preference fusion features, and perform intent analysis on the preference fusion features to obtain the user's composite cultural preferences; In this embodiment of the invention, the step of semantically aligning the cultural imagery color tags with the explicit category preference tags to obtain the user's preference fusion features includes: Semantic elements are extracted from the cultural imagery color tags and the explicit category preference tags to obtain the user's color semantic features and category semantic features; Based on the color semantic features, the category semantic features are matched with the same domain semantic features to obtain the semantic matching representation of the user; Based on the semantic matching representation, the color semantic features and the category semantic features are fused and encoded to obtain the user's preference fusion features.

[0046] For the generated cultural imagery color tags, semantic deconstruction is performed in the cultural and creative fields. The core semantic elements that carry cultural connotations and color attributes in the tags are extracted. These elements are semantic units that can directly reflect the cultural aesthetics and emotional tendencies corresponding to the colors. At the same time, the explicit category preference tags are deconstructed hierarchically. Based on the category attribute division of the cultural and creative category dictionary, the core semantic elements that represent the category type, category characteristics, and category usage attributes of cultural and creative products in the tags are extracted. The core semantic elements of the color category are integrated and encoded to form structured color semantic features. The core semantic elements of the category are integrated and encoded to form structured category semantic features. Both the color semantic features and the category semantic features retain the semantic connotations of the original tags and are presented in a unified semantic expression form to ensure the consistency of subsequent semantic analysis.

[0047] Using the cultural and creative industry as a unified semantic domain, association rules for color semantics and category semantics are established within this domain. Based on these rules, each core semantic element in the color semantic features is matched one by one with all core semantic elements in the category semantic features to determine the semantic fit between color semantic elements and category semantic elements in the cultural and creative industry scenario. The fit metric is calibrated for each set of matched semantic elements, while preserving the matching correspondence between semantic elements. The matching results and fit calibrated values ​​of all color semantic elements and category semantic elements are integrated to form a structured data set containing semantic matching correspondence and fit metric information. This data set is the user's semantic matching representation, accurately reflecting the semantic association status of color semantic features and category semantic features in the cultural and creative industry.

[0048] Based on semantic matching representation, color semantic features and category semantic features are fused. Semantic elements with high fit values ​​in the representation are prioritized for fusion, preserving their core semantic connotations. Semantic elements with low fit are retained independently, maintaining their original semantic features. During the fusion process, the semantic expression logic of the cultural and creative fields is followed, and all semantic elements are arranged hierarchically according to the tightness of semantic association. At the same time, a unified coding rule is adopted to standardize the coding of the fused semantic elements and the independently retained semantic elements. The coding results fully preserve all the core information of color semantics and category semantics, as well as the semantic matching relationship between the two. The final standardized coding results are integrated to form a structured and integrated feature set, which is the user's preference fusion feature.

[0049] The beneficial effects are as follows: by extracting semantic elements of cultural imagery color tags and explicit category preference tags in the field of cultural and creative business recommendations, the core semantics of the two types of tags are accurately decomposed and presented in a structured manner. The semantic matching completed based on the semantic matching rules of the cultural and creative field makes the correlation analysis of color and category semantics more in line with the actual business scenario of cultural and creative product recommendations, avoiding the deviation of cross-domain semantic matching. Furthermore, the fusion encoding based on semantic matching representation not only retains the core connotation of color and category semantics, but also achieves the structured integration of the two. The generated preference fusion features can accurately and comprehensively reflect the user's need for the integration of color culture preferences and category preferences in the selection of cultural and creative products. This provides unified feature data that is highly in line with the cultural and creative business recommendation scenario for the subsequent mining of users' composite cultural preferences, greatly improving the accuracy and targeting of subsequent user demand analysis and adapting to the business application needs of personalized recommendations for cultural and creative products.

[0050] Step e: Based on the aforementioned composite cultural preferences, perform an initial screening and ranking of the user's recommended cultural and creative products to obtain the user's preliminary recommended products; In this embodiment of the invention, the intention analysis performed on the preference fusion features to obtain the user's composite cultural preferences includes: By performing cultural and creative semantic intent mining on the aforementioned preference fusion features, the cultural intention representation of the user can be obtained. Based on the cultural intention representation, the cultural attribute association of the user's cultural and creative needs is determined to obtain the user's cultural attribute association representation. Based on the cultural attribute association representation, the user's cultural and creative preferences are constructed in an integrated manner to obtain the user's composite cultural preferences.

[0051] The preliminary screening and ranking of the user's cultural and creative products based on the aforementioned composite cultural preferences, to obtain the user's initial recommended products, includes: Based on the composite cultural preference profile, the cultural attribute matching search is performed on the cultural and creative products to be recommended to the user to obtain the matching degree of the user's cultural and creative products. Based on the matching degree of the cultural and creative products, the cultural and creative products to be recommended are initially screened for suitability to obtain the user's initial screened cultural and creative products. The cultural and creative products initially screened are ranked according to their cultural preference compatibility to obtain the user's preliminary recommended products.

[0052] Based on the exclusive semantic parsing rules in the field of cultural and creative business recommendations, this study deconstructs and analyzes all structured encoded semantic elements in the preference fusion features, uncovering the core intentions of users' cultural and creative consumption behind each semantic element. This includes users' cultural style preferences, color culture aesthetic demands, category selection preferences, and the tendency to express various needs in a fusion manner. The uncovered semantic intentions of cultural and creative products are classified and integrated according to the dimensions of cultural expression, category needs, and color preferences. At the same time, based on the weight ratio of semantic elements in the preference fusion features, the attention intensity of various semantic intentions of cultural and creative products is quantitatively calibrated. The classified and integrated semantic intentions of cultural and creative products are combined with the corresponding attention intensity calibration results to form a structured data set that can comprehensively and accurately reflect users' cultural needs and tendencies in cultural and creative consumption. This set is the user's cultural intention representation.

[0053] A comprehensive cultural attribute system for the cultural and creative industry is pre-established. This system fully covers multiple dimensions of cultural attribute categories, including regional culture, traditional folk customs, classic art styles, contemporary aesthetic expressions, and intangible cultural heritage connotations of cultural and creative products. Each category contains specific matching cultural attribute features. Using cultural intention representation as the core reference, the semantic intent of various cultural and creative products is matched with all cultural attribute features in the cultural attribute system one by one to determine the degree of intrinsic correlation between users' cultural and creative needs and each cultural attribute feature. The degree of correlation is quantified for each successfully matched cultural attribute feature. At the same time, the core cultural attribute category and auxiliary cultural attribute category corresponding to users' cultural and creative needs are clearly distinguished. All correlation matching results, correlation degree quantification values, and cultural attribute category classification information are systematically integrated to form a structured set of correlation judgment results. This set is the user's cultural attribute correlation representation.

[0054] Using the quantitative value of the degree of association in the cultural attribute association representation as the core weighting basis, core cultural attribute features with high degree of association values ​​are extracted. At the same time, the semantic intent of cultural and creative products with high attention intensity values ​​in the cultural intention representation is deeply integrated. Combined with the core features in the cultural image color tags and explicit category preference tags generated by users in the early stage, and according to the actual needs of personalized recommendations of cultural and creative products, these core cultural attribute features, semantic intent of cultural and creative products, and preference tag features are integrated. In the integration process, the inherent logic of the relationship between cultural attributes and consumption preferences is strictly followed. Various features are hierarchically constructed according to the degree of core and auxiliary, and the core preference dimension and auxiliary preference dimension of users' cultural and creative consumption are clearly defined. Finally, an integrated preference system is formed that includes cultural attribute tendencies, color culture preferences, and category consumption preferences, and the features of each dimension are deeply integrated and interconnected. This system is the user's composite cultural preference, which can completely and accurately reflect the integration characteristics of the user's deep cultural needs and explicit consumption preferences in cultural and creative consumption.

[0055] Using the constructed user composite cultural preferences as the core search profile, the system extracts cultural attribute features, color culture preference features, and category preference features and integrates them into a standardized search feature set. For all cultural and creative products to be recommended in the database, the system extracts the core features of each product, such as cultural attributes, color design connotations, category affiliation, and cultural style expression. The extracted product features are then matched and compared with the user's search feature set in a full-dimensional and multi-dimensional manner. According to the pre-set multi-dimensional matching rules for cultural and creative products, the matching results of each feature are refined and quantitatively scored. Then, corresponding weights are assigned according to the importance ratio of each feature in the recommendation of cultural and creative products. The quantitative scores of each feature are multiplied by their corresponding weights and summed. The resulting comprehensive value is the matching degree of the cultural and creative product, accurately quantifying the degree of fit between the product and the user's composite cultural preferences.

[0056] To meet the commercial application needs of personalized recommendations for cultural and creative products, a product suitability screening standard tailored to actual recommendation scenarios is pre-defined. This standard is based on the core dimension of users' complex cultural preferences. It iterates through the matching scores of all cultural and creative products to be recommended in the database. Based on the established suitability screening standard, products whose matching scores align with users' core cultural preferences and meet the suitability requirements are selected. Products whose matching scores do not meet the suitability requirements or whose fit with users' core preferences is low are eliminated. All selected cultural and creative products that meet the suitability requirements are then compiled into a product set. This set represents the user's initial screening of cultural and creative products, ensuring that all screened products have a basic and effective fit with the user's complex cultural preferences.

[0057] Using the matching score of each initially screened cultural and creative product as the core quantitative indicator of cultural preference fit, all initially screened cultural and creative products are systematically sorted in descending order of matching score, with the product having the highest cultural preference fit at the top, and products with decreasing fit arranged sequentially to the bottom. The sorting process strictly follows the numerical ranking rules to ensure the accuracy and objectivity of the sorting results. The sequence of initially screened cultural and creative products after descending sorting is taken as a whole, which is the user's initial recommended products. This ensures that the products in the sequence are presented hierarchically according to their degree of fit with the user's cultural preferences, laying the foundation for subsequent refined screening and matching.

[0058] The beneficial effects are as follows: by mining the semantic intent of cultural and creative products based on the fusion features of preferences and determining the association with cultural attributes, it achieves precise mining from the fusion preference features of users to their deep cultural needs. This allows the constructed composite cultural preferences to fully match the explicit needs and deep cultural aesthetic tendencies of users' cultural and creative consumption, providing a highly accurate user profile basis for product recommendations. Subsequently, based on the cultural attribute matching retrieval, adaptability screening, and fit ranking of composite cultural preferences, the degree of fit between products and user preferences is quantified from multiple dimensions. Through a standardized screening and ranking process, precise initial screening of cultural and creative products to be recommended is achieved. This allows the initially recommended products to be presented hierarchically according to their degree of fit with users' cultural preferences. This not only ensures the basic adaptability of the initially screened products to users' core preferences, but also defines the precise range for subsequent refined visual feature matching screening, greatly improving the efficiency and accuracy of the cultural and creative product recommendation process and highly adapting to the actual application needs in the field of cultural and creative commercial recommendation.

[0059] Step f: Perform product image segmentation on the initially recommended products to obtain the pattern shape code of the initially recommended products, and perform collaborative screening and matching of the pattern shape code with the cultural image color tag to obtain the personalized recommended products for the user.

[0060] In this embodiment of the invention, the step of performing product image segmentation on the initially recommended products to obtain the pattern shape code of the initially recommended products, and then performing collaborative screening and matching of the pattern shape code with the cultural imagery color tag to obtain the personalized recommended products for the user, includes: Perform foreground-background separation processing on the product image of the initially recommended product to obtain the main pattern image of the initially recommended product; The outline features of the pattern main image are extracted to obtain the pattern outline features of the preliminary recommended product; The pattern outline features are structured and encoded to obtain the pattern shape code of the preliminary recommended product; Calculate the degree of coordination between the pattern shape code and the cultural image color tag; Based on the collaborative matching degree, the initially recommended products are collaboratively screened to obtain personalized recommended products for the user.

[0061] The formula for calculating the collaborative matching degree is as follows: ; in, The degree of collaborative matching, The feature value encoded by the pattern shape, The feature value of the cultural image color tag, The category weight of the explicit category preference label. The cultural preference weights for the aforementioned cultural image color labels. This represents the domain correlation coefficient between pattern and color.

[0062] High-resolution product images of the initially recommended products are obtained. With the goal of identifying patterns in cultural and creative products, a threshold for distinguishing image pixels is set. By analyzing the color saturation, brightness, and pixel neighborhood features of the image pixel by pixel, the pixel areas belonging to the product's own pattern are determined as foreground areas, while the pixel areas in the product image that are not pattern elements, such as background layout and lighting, are determined as background areas. Through precise division of pixel areas, the foreground and background are completely separated. After removing the pixel information of all background areas, the complete foreground pattern pixel areas are retained. The image composed of the pure pattern pixel areas is the main pattern image of the initially recommended product, ensuring that the image contains only the core pattern information of the product and no other interfering elements.

[0063] Pixel edge detection is performed on the main image of the pattern, identifying edge pixels where the color and brightness of pixels change abruptly row by row and column by column. All continuous edge pixels are connected and integrated to form the complete outer contour and internal detail contour lines of the main body of the pattern. The integrated contour lines are smoothed to eliminate pixel-level contour burrs and irregular breaks, while retaining the core geometric shape, line direction, pattern structure and detail texture features of the pattern contour. These feature information that can accurately reflect the essence of the pattern shape are summarized to form a structured feature set. This set is the pattern contour feature of the initially recommended product, which fully restores the shape and structural characteristics of the product pattern.

[0064] A standardized coding rule for the pattern outline of cultural and creative products is established. This rule sets exclusive coding dimensions for the geometric shape, line curvature, pattern symmetry, and structural hierarchy of the pattern outline. The extracted pattern outline features are split into dimensions according to the coding rule. The feature information of each dimension is digitally converted and symbolically represented. The coding results of each dimension are concatenated and combined according to the preset coding order to form a unique coding sequence that can reversely restore the pattern outline features. This coding sequence is the pattern shape code of the initially recommended product, realizing the standardization, digital storage and comparison of pattern outline features.

[0065] Collaborative matching degree is used to quantify the degree of fit between the pattern shape features of cultural and creative products and the user's cultural imagery color preferences. By combining the pattern shape encoding feature value and the cultural imagery color label feature value with their corresponding weights and performing a weighted square operation, and then performing a weighted sum with the result of the harmonic mean of the two, a value that can simultaneously reflect the degree of matching between the pattern shape and the cultural imagery color is obtained, which serves as a quantitative indicator of the degree of collaborative matching between cultural and creative products and user preferences.

[0066] The parameters for calculating the collaborative matching degree all come from the feature extraction and weight allocation process of cultural and creative products and user preferences. The feature values ​​of the pattern shape encoding are obtained by segmenting the product image to extract the pattern outline and encoding it. The feature values ​​of the cultural image color tag are obtained by performing color analysis on the visual focus area of ​​the user's browsing trajectory and generating tag encoding. The category weight of the explicit category preference tag is obtained by performing word frequency statistics and normalization on the user's historical order data. The cultural preference weight of the cultural image color tag is obtained by performing frequency statistics and normalization on the user's cultural image preference tag. The domain correlation coefficient of pattern and color is obtained by statistically analyzing and assigning values ​​to the correlation between pattern features and color features in the cultural and creative product domain.

[0067] The degree of collaborative matching increases monotonically with the increase of the feature values ​​of pattern shape encoding and cultural image color label. When the feature values ​​of pattern shape encoding and cultural image color label increase simultaneously, the weighted square operation term and the harmonic mean operation term also increase synchronously, and the degree of collaborative matching after the weighted sum of the two also increases. When the domain correlation coefficient between pattern and color increases, the proportion of the weighted square operation term in the calculation of collaborative matching degree increases. When the domain correlation coefficient decreases, the proportion of the harmonic mean operation term in the calculation of collaborative matching degree increases. The change in collaborative matching degree always maintains a positive correlation with the two feature values.

[0068] A screening threshold for the synergy matching degree of cultural and creative products is set. This threshold is preset based on the accuracy requirements of personalized recommendations for cultural and creative products. The synergy matching degree values ​​of all initially recommended products are traversed, and the initially recommended products with matching degree values ​​higher than the screening threshold are filtered out. The filtered products are then sorted in descending order of synergy matching degree values. If the number of filtered products exceeds the preset recommendation quantity range, the preset number of products at the top of the ranking are selected. The finally filtered and sorted cultural and creative products are summarized to form a product set. This set is the user's personalized recommended products, ensuring that the recommended products are highly consistent with the user's preferences in terms of pattern shape and cultural imagery color.

[0069] The beneficial effects are as follows: by performing refined foreground and background separation on the initial recommended product images, the pure pattern subject image is accurately extracted, laying an interference-free image foundation for subsequent pattern feature analysis. Then, through contour feature extraction and structured encoding, the digital and standardized transformation of pattern shape features is realized, giving pattern features quantifiable and comparable attributes. Combined with cultural imagery color tags to calculate the collaborative matching degree, the precise quantitative matching of product features and user preferences is achieved from the dual dimensions of pattern and color. Finally, based on the matching degree, collaborative screening and ranking further improve the fit between recommended products and users' deep cultural aesthetic preferences. The entire process realizes the standardized processing of the entire chain from product visual feature extraction to precise matching and screening, so that the final personalized recommended products not only fit the user's category and color preferences, but also match the user's pattern aesthetic needs, greatly improving the accuracy and personalization of personalized recommendations for cultural and creative products, and adapting to the actual needs of commercial recommendations for cultural and creative products.

[0070] like Figure 2 The diagram shown is a functional module diagram of a personalized recommendation system for cultural and creative products based on user profiles, provided in an embodiment of the present invention.

[0071] The personalized recommendation system 100 for cultural and creative products based on user profiles described in this invention can be installed in an electronic device. Depending on the functions implemented, the personalized recommendation system 100 may include a visual focus area extraction module 101, a cultural imagery color tag generation module 102, an explicit category preference tag determination module 103, a composite cultural preference construction module 104, a preliminary recommended product screening module 105, and a personalized product recommendation module 106. The modules described in this invention can also be referred to as units, which are a series of computer program segments that can be executed by the processor of an electronic device and perform a fixed function, stored in the memory of the electronic device.

[0072] In this embodiment, the functions of each module / unit are as follows: The visual focus area extraction module 101 is used to collect the user's browsing trajectory data and mark the dwell time of the coordinate points in the browsing trajectory data to obtain the user's visual focus area. The cultural imagery color tag generation module 102 is used to perform pixel color analysis on the visual focus area to obtain the user's color distribution histogram, and to perform peak detection on the color distribution histogram to obtain the user's cultural imagery color tag. The explicit category preference tag determination module 103 is used to perform word frequency statistics on the user's historical order data to obtain the user's explicit category preference tags; The composite cultural preference construction module 104 is used to semantically align the cultural image color tags with the explicit category preference tags to obtain the user's preference fusion features, and to perform intent analysis on the preference fusion features to obtain the user's composite cultural preferences. The preliminary recommended product screening module 105 is used to perform preliminary screening and sorting of the user's cultural and creative products based on the composite cultural preferences, so as to obtain the user's preliminary recommended products; The personalized product recommendation module 106 is used to perform product image segmentation on the initially recommended products to obtain the pattern shape code of the initially recommended products, and to perform collaborative screening and matching of the pattern shape code with the cultural image color tag to obtain the personalized recommended products for the user.

[0073] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0074] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0075] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0076] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0077] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0078] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A personalized recommendation method for cultural and creative products based on user profiles, characterized in that, The method includes: Step a: Collect the user's browsing trajectory data and mark the dwell time of the coordinate points in the browsing trajectory data to obtain the user's visual focus area; Step b: Perform pixel color analysis on the visual focus area to obtain the user's color distribution histogram, and perform peak detection on the color distribution histogram to obtain the user's cultural image color tag; Step c: Perform word frequency statistics on the user's historical order data to obtain the user's explicit category preference tags; Step d: Semantically align the cultural image color tags with the explicit category preference tags to obtain the user's preference fusion features, and perform intent analysis on the preference fusion features to obtain the user's composite cultural preferences; Step e: Based on the aforementioned composite cultural preferences, perform an initial screening and ranking of the user's recommended cultural and creative products to obtain the user's preliminary recommended products; Step f: Perform product image segmentation on the initially recommended products to obtain the pattern shape code of the initially recommended products, and perform collaborative screening and matching of the pattern shape code with the cultural image color tag to obtain the personalized recommended products for the user.

2. The personalized recommendation method for cultural and creative products based on user profiles as described in claim 1, characterized in that, The process of collecting user browsing trajectory data and marking the dwell time of coordinate points in the browsing trajectory data to obtain the user's visual focus area includes: Obtain the user's browsing interface coordinate stream data, and remove noise points from the browsing interface coordinate stream data to obtain the user's valid browsing trajectory point sequence; The time stamps of the trajectory points in the valid browsing trajectory point sequence are parsed to obtain the spatiotemporal dwell segment of the user; The coordinates of the first and last trajectory points of the spatiotemporal dwell segment are merged to obtain the gaze center point of the spatiotemporal dwell segment; The duration of the gaze center point is determined based on the number of trajectory points within the spatiotemporal dwell segment, the first trajectory point, and the last trajectory point. The duration of the gaze is assigned as attribute information to the gaze center point to obtain the user's visual focus area.

3. The personalized recommendation method for cultural and creative products based on user profiles as described in claim 1, characterized in that, The step involves performing pixel color analysis on the visual focus area to obtain the user's color distribution histogram, and then performing peak detection on the color distribution histogram to obtain the user's cultural imagery color tag, including: Extract the gaze center point in the visual focus area, and define a color sampling window based on the gaze center point; Collect the pixel color channel values ​​in the color sampling window, and deduplicate and merge the pixel color channel values ​​to obtain the user's gaze point color features; The color features of the gaze point are frequency-collected to obtain the color frequency distribution histogram of the user. Peak values ​​are extracted from the frequency distribution histogram of the color system to obtain the user's significant color system categories; The significant color categories are associated and matched with a preset cultural imagery mapping library to obtain the user's cultural imagery color tags.

4. The personalized recommendation method for cultural and creative products based on user profiles as described in claim 1, characterized in that, The step of performing word frequency statistics on the user's historical order data to obtain the user's explicit category preference tags includes: Extract and parse the user's historical order data to obtain the product title text of the historical order data; Stop words are removed from the product title text to obtain the user's product keyword sequence; The product keyword sequence is matched and mapped with a preset cultural and creative product category dictionary to obtain the user's effective category keywords; The frequency of the effective category keywords in the product keyword sequence is counted to obtain the original frequency of the effective category keywords; Based on the original frequency, a time decay coefficient is assigned to the effective category keywords to obtain the weighted time-effect frequency of the effective category keywords; Based on the weighted timeliness frequency, the effective category keywords are sorted and filtered to obtain the user's high-frequency core category keywords; The high-frequency core category words are associated and aligned with the category hierarchy structure in the cultural and creative category dictionary to obtain the user's explicit category preference tags.

5. The personalized recommendation method for cultural and creative products based on user profiles as described in claim 1, characterized in that, The semantic alignment of the cultural imagery color tags with the explicit category preference tags to obtain the user's preference fusion features includes: Semantic elements are extracted from the cultural imagery color tags and the explicit category preference tags to obtain the user's color semantic features and category semantic features; Based on the color semantic features, the category semantic features are matched with the same domain semantic features to obtain the semantic matching representation of the user; Based on the semantic matching representation, the color semantic features and the category semantic features are fused and encoded to obtain the user's preference fusion features.

6. The personalized recommendation method for cultural and creative products based on user profiles as described in claim 1, characterized in that, The intention analysis is performed on the preference fusion features to obtain the user's composite cultural preferences, including: By performing cultural and creative semantic intent mining on the aforementioned preference fusion features, the cultural intention representation of the user can be obtained. Based on the cultural intention representation, the cultural attribute association of the user's cultural and creative needs is determined to obtain the user's cultural attribute association representation. Based on the cultural attribute association representation, the user's cultural and creative preferences are constructed in an integrated manner to obtain the user's composite cultural preferences.

7. The personalized recommendation method for cultural and creative products based on user profiles as described in claim 1, characterized in that, The preliminary screening and ranking of the user's cultural and creative products based on the aforementioned composite cultural preferences, to obtain the user's initial recommended products, includes: Based on the composite cultural preference profile, the cultural attribute matching search is performed on the cultural and creative products to be recommended to the user to obtain the matching degree of the user's cultural and creative products. Based on the matching degree of the cultural and creative products, the cultural and creative products to be recommended are initially screened for suitability to obtain the user's initial screened cultural and creative products. The cultural and creative products initially screened are ranked according to their cultural preference compatibility to obtain the user's preliminary recommended products.

8. The personalized recommendation method for cultural and creative products based on user profiles as described in claim 1, characterized in that, The process involves segmenting the product images of the initially recommended products to obtain their pattern shape codes, and then performing collaborative filtering and matching of these pattern shape codes with the cultural imagery color tags to obtain the user's personalized product recommendations, including: Perform foreground-background separation processing on the product image of the initially recommended product to obtain the main pattern image of the initially recommended product; The outline features of the pattern main image are extracted to obtain the pattern outline features of the preliminary recommended product; The pattern outline features are structured and encoded to obtain the pattern shape code of the preliminary recommended product; Calculate the degree of coordination between the pattern shape code and the cultural image color tag; Based on the collaborative matching degree, the initially recommended products are collaboratively screened to obtain personalized recommended products for the user.

9. The personalized recommendation method for cultural and creative products based on user profiles as described in claim 8, characterized in that, The formula for calculating the collaborative matching degree is as follows: ； in, The degree of collaborative matching, The feature value encoded by the pattern shape, The feature value of the cultural image color tag, The category weight of the explicit category preference label. The cultural preference weights for the aforementioned cultural image color labels. This represents the domain correlation coefficient between pattern and color.

10. A personalized recommendation system for cultural and creative products based on user profiles, characterized in that, The system for implementing the personalized recommendation method for cultural and creative products based on user profiles as described in claim 1 includes: The visual focus area extraction module is used to collect the user's browsing trajectory data and mark the dwell time of the coordinate points in the browsing trajectory data to obtain the user's visual focus area. The cultural imagery color tag generation module is used to perform pixel color analysis on the visual focus area to obtain the user's color distribution histogram, and to perform peak detection on the color distribution histogram to obtain the user's cultural imagery color tag. The explicit category preference tag determination module is used to perform word frequency statistics on the user's historical order data to obtain the user's explicit category preference tags; The composite cultural preference construction module is used to semantically align the cultural image color tags with the explicit category preference tags to obtain the user's preference fusion features, and to perform intent analysis on the preference fusion features to obtain the user's composite cultural preferences. The preliminary recommended product screening module is used to perform preliminary screening and sorting of the user's cultural and creative products based on the composite cultural preferences, so as to obtain the user's preliminary recommended products; The personalized product recommendation module is used to perform product image segmentation on the initially recommended products to obtain the pattern shape code of the initially recommended products, and to perform collaborative screening and matching of the pattern shape code with the cultural image color tag to obtain the personalized recommended products for the user.

Images