Intelligent movie content recommendation method and system based on artificial intelligence

Abstract

The application discloses a movie and television content intelligent recommendation method and system based on artificial intelligence, and the method comprises the steps of collecting multiple data sources, optimizing data, recommending group movie and television content, optimizing and sorting personalized recommendation, and intelligently recommending movie and television content. The application relates to the technical field of data processing, and particularly relates to a movie and television content intelligent recommendation method and system based on artificial intelligence. The application innovatively combines group recommendation and personalized optimization, realizes dynamic balance between group representation and individual difference, innovatively proposes a density correction mechanism, weighted distance calculation, representative point merging mechanism and fuzzy weighted distribution strategy to improve a clustering algorithm, improves clustering accuracy, and improves the accuracy of group recommendation and user satisfaction. An adaptive reference individual selection strategy and an iterative global exploration weight-based optimization algorithm are used to significantly improve the accuracy of model output results, and the stability of personalized movie and television content recommendation optimization sorting is realized.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to an intelligent recommendation method and system for film and television content based on artificial intelligence. Background Technology

[0002] The AI-based intelligent recommendation method and system for film and television content refers to a system that utilizes artificial intelligence technology to collect multi-source heterogeneous data, analyze film and television content and user behavior data, and construct a personalized and accurate recommendation system. This system realizes the transformation of film and television content distribution from traditional static display to dynamic personalized recommendation, effectively promoting the automation, accuracy and efficiency of the recommendation process, and improving user experience and content dissemination efficiency.

[0003] However, traditional film and television content recommendation methods suffer from technical problems such as relying solely on a single dimension of group or individual analysis, leading to homogenized recommendation results, difficulty in reflecting group differences, and insufficient adaptation to users' dynamic interests, thus affecting the accuracy of film and television content recommendations and user experience. Existing clustering algorithms suitable for user group segmentation tend to favor high-density clusters when dealing with datasets with significant density differences, causing sparse users to be ignored or misclassified, resulting in unstable clustering results and inaccurate group identification, thus affecting the accuracy of film and television content recommendations. Existing personalized film and television ranking models suffer from unreasonable model parameter settings, resulting in unstable personalized rating results and insufficient accuracy for film and television content. Summary of the Invention

[0004] To address the aforementioned issues and overcome the shortcomings of existing technologies, this invention provides an intelligent recommendation method and system for film and television content based on artificial intelligence. Traditional film and television content recommendation methods rely solely on a single dimension of group or individual analysis, leading to homogenized recommendation results, difficulty in reflecting group differences, and insufficient adaptation to dynamic user interests, thus affecting recommendation accuracy and user experience. This solution innovatively combines group recommendation with personalized optimization. At the group level, users are divided into multiple groups based on clustering modeling, generating a recommendation set that reflects group characteristics. At the individual level, the group recommendation results are optimized and ranked by combining user personality and contextual data, achieving a dynamic balance between group representativeness and individual differences. This effectively overcomes the shortcomings of traditional recommendation methods in the disconnect between group and individual modeling, improving the diversity and personalization of recommendation results, and ultimately achieving a comprehensive improvement in user satisfaction, activity, and overall film and television content recommendation accuracy. Furthermore, existing clustering algorithms applicable to user group segmentation tend to favor high-density clusters when handling datasets with significant density differences, leading to the neglect or misclassification of sparse user groups, resulting in unstable clustering results and group... To address the technical problem of inaccurate identification affecting the accuracy of film and television content recommendations, this solution innovatively introduces a density correction mechanism, weighted distance calculation, representative point merging mechanism, and fuzzy weighted allocation strategy to improve the clustering algorithm. This effectively avoids the omission of sparse groups, ensures the accuracy of cluster center selection, and reduces the interference of outliers and noisy data on clustering results, thereby improving the overall stability and reliability of clustering. It can more accurately characterize the user group structure, thus significantly improving the accuracy and diversity of film and television content recommendations, and enhancing the accuracy of group film and television content recommendations in complex user data environments. To address the technical problem of unreasonable model parameter settings in existing personalized film and television ranking models, leading to unstable and inaccurate personalized rating results, this solution adopts an adaptive reference individual selection strategy and an iterative global exploration weight improvement optimization algorithm. It performs global and local joint search on key parameters in the personalized film and television ranking model, automatically obtaining the optimal model hyperparameter combination, effectively improving the intelligence and accuracy of hyperparameter search, significantly improving the accuracy and stability of model output results, and achieving stability in personalized film and television content recommendation optimization ranking.

[0005] The technical solution adopted by this invention is as follows: The intelligent recommendation method and system for film and television content based on artificial intelligence provided by this invention includes the following steps:

[0006] Step S1: Multi-source data acquisition;

[0007] Step S2: Data optimization;

[0008] Step S3: Group film and television content recommendation;

[0009] Step S4: Personalized recommendation and sorting optimization;

[0010] Step S5: Intelligent recommendation of film and television content.

[0011] Further, in step S1, the multi-source data collection specifically involves collecting data through a film and television content platform to obtain raw film and television recommendation data. The raw film and television recommendation data includes historical film and television recommendation data, real-time film and television recommendation data, reference personalized ranking data, and target personalized ranking data. Both the historical film and television recommendation data and the real-time film and television recommendation data include user basic information data, user behavior data, and user preference data. The reference personalized ranking data and the target personalized ranking data include user basic information data, user behavior data, film and television content data, and recommendation environment data.

[0012] Furthermore, in step S2, the data optimization specifically includes the following steps:

[0013] Step S21: Data cleaning, specifically, filling in missing values, removing outliers, and processing duplicates in the original data;

[0014] Step S22: Data format conversion, specifically time field conversion, text field conversion, and behavioral sequence data conversion;

[0015] Step S23: Data encoding processing, specifically, using one-hot encoding to encode the discrete and categorical variables in the original data, performing numerical processing, and converting them into sparse binary vector form;

[0016] Step S24: Selecting film and television recommendation features. Specifically, the correlation coefficient analysis method is used to extract feature variables related to film and television content recommendations from the original data to form a set of recommendation features.

[0017] Furthermore, in step S3, the group's film and television content recommendation specifically includes the following steps:

[0018] Step S31: Construct a clustering algorithm for user group segmentation, which specifically includes the following steps:

[0019] Step S311: Similarity calculation, specifically using Euclidean distance to calculate the similarity between data points;

[0020] Step S312: Corrected density calculation, specifically, involves calculating the variability density by introducing a density correction factor to obtain the corrected density value; the formula used is as follows:

[0021] ;

[0022] In the formula, This represents the maximum local density. This represents the corrected density value for data point i. Indicates the cutoff distance threshold. This represents the density correction factor, with a range of values. Used to adjust the weights of sparse populations. This represents the similarity value between data point i and data point j;

[0023] Step S313: Weighted distance calculation. Specifically, when calculating the minimum distance to higher density points, dynamic density weights are introduced for correction to obtain the weighted distance value; the formula used is as follows:

[0024] ;

[0025] In the formula, This represents the minimum distance from data point i to data points with a higher density. This represents the weighted distance to data point i;

[0026] Step S314: Cluster center formation, specifically, firstly, a set of candidate representative points is selected based on both modified density and weighted distance conditions. Then, the candidate representative points are merged using the representative point merging threshold to finally generate a stable set of cluster centers.

[0027] The merging operation specifically involves calculating the similarity between any two data points in the candidate representative point set. If the similarity is... If these two representative points are too close and belong to the same user group, they should be merged, and the candidate representative point with the higher modified density should be retained as the new cluster center. If the cluster centers are different, then they are retained as different cluster centers; the formula used is as follows:

[0028] ;

[0029] In the formula, This represents the threshold for merging data points, where n represents the number of data points. This represents the adjustment factor, used to control the threshold size;

[0030] Step S315: Initial micro-cluster construction, specifically, involves detecting outliers in the data points using the local anomaly factor method, identifying normal points that are close to the cluster center and are not anomalous as candidate assignment objects, and directly assigning normal points to the group with the same nearest cluster center, thereby forming local initial micro-clusters; the formula used is as follows:

[0031] ;

[0032] In the formula, Represents the set of normal points. Represents the local anomaly factor function. The vector representing the i-th data point;

[0033] Step S316: Residual point reallocation, specifically including membership calculation, optimal cluster selection, and iterative allocation of residual points; including the following steps:

[0034] Step S3161: Membership degree calculation, specifically, for each unassigned data point, calculate the membership degree value of each data point to each cluster based on the dynamic fuzzy weighting method; the formula used is as follows:

[0035] ;

[0036] ;

[0037] In the formula, This represents the similarity weight between data point i and data point j. This represents the nearest neighbor criterion weight between data point i and data point j. This represents the similarity weight between data point e and data point j. This represents the membership value of data point i to cluster c. This represents the set of k nearest neighbors of data point i. This represents the set of k nearest neighbors of data point j. This indicates the cluster to which data point j belongs;

[0038] Step S3162: Optimal cluster selection, specifically for each unassigned data point, recording its maximum membership value and the corresponding cluster;

[0039] Step S3163: Iterative allocation of remaining points. Specifically, select the data point p with the highest membership degree among all unallocated data points, allocate it to the corresponding cluster, and update its allocation status. At the same time, update the membership degree of the k nearest neighbor data points q of data point p to each cluster, and recalculate the new maximum membership degree value and the corresponding cluster label. Repeat the above allocation process until all remaining points are allocated.

[0040] Step S317: Obtain clustering results. Specifically, after all data points have been assigned, they are aggregated into the corresponding cluster labels to form multiple independent clusters, each cluster corresponding to a user group with similar characteristics.

[0041] Step S32: Obtain real-time user clustering results. Specifically, input the historical film and television recommendation data and the real-time film and television recommendation data in the film and television recommendation optimization data into the clustering algorithm for user group division, generate multiple independent user clusters, evaluate the clustering quality of each cluster, calculate its corresponding silhouette coefficient, and filter out the clusters with silhouette coefficients higher than a preset threshold to obtain the user group division results.

[0042] Step S33: Obtain the set of recommended film and television content for the target user. Specifically, based on the user group segmentation results, determine the group to which the target user belongs, statistically analyze and quantify the historical recommendation feedback satisfaction of the film and television content within the group, filter out the film and television content with excellent and high feedback levels within the group, and combine them to form the set of recommended film and television content for the target user.

[0043] Furthermore, in step S4, the personalized recommendation optimization ranking is used to perform personalized user rating and ranking on the group-recommended film and television content set; specifically, it includes the following steps:

[0044] Step S41: Construct and train a personalized film and television ranking model. Specifically, a personalized film and television ranking model is established based on a deep neural network. Reference personalized ranking data is used as the training data for the model. The model is trained to obtain the trained personalized film and television ranking model.

[0045] Step S42: Model hyperparameter optimization, specifically, using an improved optimization algorithm to optimize the hyperparameters in the model to obtain the optimal combination of model hyperparameters; this includes the following steps:

[0046] Step S421: Initialize the search population individuals. Specifically, through random initialization, generate the position vectors of N search individuals in the parameter space. Each individual encodes a candidate model hyperparameter combination to form the initial search population.

[0047] Step S422: Calculate the individual fitness value, specifically by calculating the fitness value of the search individual in the population; and use the performance of the trained personalized film and television ranking model based on the search individual's position as the individual's fitness value.

[0048] Step S423: Adaptive reference individual selection, used to dynamically select the location of the interpolation point based on the individual's fitness value and its distribution in the search space, specifically for each search individual. Two random search individuals and The three individuals are ranked according to their fitness values ​​to determine the best individual. Name the remaining two individuals and Based on the positional relationship of the three individuals in the search space, the position of the adaptive reference individual is calculated. The formula used is as follows:

[0049] ;

[0050] In the formula, This represents a positive search factor, used to explore small areas on either side of the current individual. This represents a reverse search factor, used to extend the reference individual's location outwards;

[0051] Step S424: Perform a global exploration, specifically if Then, a global exploration operation is performed, using the adaptive reference individual position to guide the current individual's position update; the formula used is as follows:

[0052] ;

[0053] ;

[0054] In the formula, Represents the global exploration weight in the t-th iteration. This represents the scaling factor, with a value range of [0.2, 0.8]. , , and Both represent random numbers uniformly distributed within the range [0,1], and t represents the current iteration number. Indicates the maximum number of iterations. Indicates a random search for individuals. This represents the rounding function. Represents the natural logarithm function. This represents the position of the in-th search individual in the t-th iteration;

[0055] Step S425: Perform local exploration, specifically if If the current optimal individual position is used, a local exploration operation is performed to guide the position update of the current iteration individual.

[0056] Step S426: Obtaining the optimal position of an individual. Specifically, after each iteration, the fitness values ​​of all individuals in the current population are evaluated. If the fitness of an individual's position is better than that of the current global optimal individual position, then the global optimal individual position is updated using that individual.

[0057] Step S427: Search iteration terminates, specifically when the fitness value of the search individual is higher than the fitness threshold and the maximum number of iterations is reached, the search is terminated and the global optimal individual position is obtained. The global optimal individual position specifically refers to the optimal combination of model hyperparameters of the personalized film and television ranking model.

[0058] Step S43: Obtain the optimal personalized film and television ranking model. Specifically, based on the optimal model hyperparameter combination, adjust the hyperparameters of the trained personalized film and television ranking model to obtain the optimal personalized film and television ranking model.

[0059] Step S44: Personalized film and television content recommendation and ranking, specifically, inputting the target personalized ranking data into the optimal personalized film and television ranking model, generating personalized rating results for each film and television content in the group recommended film and television content set, and ranking the film and television content in the group recommended film and television content set according to the rating results, thereby obtaining a personalized film and television content recommendation sequence that meets the target user.

[0060] Furthermore, in step S5, the intelligent recommendation of film and television content specifically involves generating a final list of recommended film and television content based on the personalized film and television content recommendation sequence, and displaying it on the target user's film and television recommendation interface, thereby achieving accurate matching and intelligent recommendation between the user and the film and television content.

[0061] The technical solution adopted by the present invention is as follows: The intelligent recommendation system for film and television content based on artificial intelligence provided by the present invention includes a multi-source data acquisition module, a data optimization module, a group film and television content recommendation module, a personalized recommendation optimization and sorting module, and an intelligent recommendation module for film and television content;

[0062] The multi-source data acquisition module obtains raw film and television recommendation data through data acquisition operations and sends the raw film and television recommendation data to the data optimization module;

[0063] The data optimization module receives data sent by the multi-source data acquisition module, performs data cleaning, data standardization, data encoding processing and film and television recommendation feature selection on the raw film and television recommendation data to obtain optimized film and television recommendation data, and sends the optimized film and television recommendation data to the group film and television content recommendation module and the personalized recommendation optimization ranking module.

[0064] The group film and television content recommendation module receives data sent by the data optimization module. Specifically, it constructs a clustering algorithm for user group segmentation based on an improved clustering algorithm, inputs the film and television recommendation data in the film and television recommendation optimization data into the clustering algorithm, generates user group segmentation results, combines the historical recommendation feedback information of each group, filters and combines the film and television content within the group, obtains the group recommended film and television content set, and sends the data to the personalized recommendation optimization and sorting module.

[0065] The personalized recommendation optimization and ranking module receives data sent by the data optimization module and the personalized recommendation optimization and ranking module. Specifically, it constructs and trains a personalized film and television ranking model, and obtains the optimal model hyperparameter combination by combining the improved optimization algorithm to obtain the optimal personalized film and television ranking model. The target data is input into it to generate personalized rating results for each film and television content. The film and television content is ranked according to the rating results to obtain a personalized film and television content recommendation sequence that meets the target user. The data is then sent to the film and television content intelligent recommendation module.

[0066] The intelligent film and television content recommendation module receives data from the personalized recommendation optimization and sorting module, generates a final film and television content recommendation list based on the personalized film and television content recommendation sequence, and displays it on the user's end, realizing accurate matching and intelligent push of film and television content between users.

[0067] The beneficial effects achieved by the present invention using the above solution are as follows:

[0068] (1) In view of the technical problems in traditional film and television content recommendation methods, which rely solely on the analysis of a single dimension of group or individual, resulting in homogenized recommendation results, difficulty in reflecting group differences, and insufficient adaptation to users' dynamic interests, thus affecting the accuracy of film and television content recommendation and user experience, this solution innovatively proposes to combine group recommendation with personalized optimization. At the group level, users are divided into multiple groups based on cluster modeling to generate a recommendation set that reflects the characteristics of the groups. At the individual level, the group recommendation results are optimized and sorted by combining user personality and contextual data to achieve a dynamic balance between group representativeness and individual differences. This effectively overcomes the shortcomings of traditional recommendation methods in terms of the disconnect between group and individual modeling, improves the diversity and personalization of recommendation results, and ultimately achieves a comprehensive improvement in user satisfaction, activity, and overall accuracy of film and television content recommendation.

[0069] (2) To address the technical problem that existing clustering algorithms applicable to user group segmentation tend to favor high-density clusters when processing datasets with significant density differences, leading to the neglect or misclassification of sparse user groups, resulting in unstable clustering results and inaccurate group identification, thus affecting the accuracy of film and television content recommendation, this solution innovatively introduces a density correction mechanism, weighted distance calculation, representative point merging mechanism, and fuzzy weighted allocation strategy to improve the clustering algorithm. This effectively avoids the omission of sparse groups, ensures the accuracy of cluster center selection, and reduces the interference of outliers and noisy data on the clustering results, thereby improving the overall stability and reliability of clustering. It can more accurately depict the user group structure, thus significantly improving the accuracy and diversity of film and television content recommendation, and achieving the improvement of the accuracy of group film and television content recommendation in complex user data environments.

[0070] (3) In view of the technical problems of unreasonable model parameter settings in existing personalized film and television ranking models, resulting in unstable and inaccurate personalized scoring results of film and television content, this solution adopts an adaptive reference individual selection strategy and an iterative global exploration weight improvement optimization algorithm to perform global and local joint search on the key parameters in the personalized film and television ranking model, automatically obtain the optimal model hyperparameter combination, effectively improve the intelligence and accuracy of hyperparameter search, significantly improve the accuracy and stability of model output results, and realize the stability of personalized film and television content recommendation optimization ranking. Attached Figure Description

[0071] Figure 1 A flowchart illustrating the AI-based intelligent recommendation method for film and television content provided by this invention;

[0072] Figure 2 A schematic diagram of the modules of the AI-based intelligent recommendation system for film and television content provided by the present invention;

[0073] Figure 3 This is a flowchart illustrating step S3;

[0074] Figure 4 This is a flowchart illustrating step S31;

[0075] Figure 5 This is a flowchart illustrating step S4;

[0076] Figure 6 This is a flowchart illustrating step S42;

[0077] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0078] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0079] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the system or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0080] Example 1, see Figure 1 The technical solution adopted by this invention is as follows: The intelligent recommendation method for film and television content based on artificial intelligence provided by this invention includes the following steps:

[0081] Step S1: Multi-source data collection. Through data collection operations, the raw data for film and television recommendations is obtained.

[0082] Step S2: Data optimization, which optimizes the collected raw film and television recommendation data. Specifically, it involves data cleaning, data standardization, data encoding, and film and television recommendation feature selection to obtain optimized film and television recommendation data.

[0083] Step S3: Group-based film and television content recommendation, used to recommend film and television content at the user group level. Specifically, it involves constructing a clustering algorithm for user group segmentation based on an improved clustering algorithm, inputting film and television recommendation data from the film and television recommendation optimization data into the clustering algorithm, generating user group segmentation results, and combining historical recommendation feedback information of each group to filter and combine film and television content within the group to obtain a set of recommended film and television content for the group.

[0084] Step S4: Personalized recommendation optimization and ranking, used to optimize the group recommendation ranking of film and television content at the individual user level. Specifically, it involves building and training a personalized film and television ranking model, and combining it with an improved optimization algorithm to obtain the optimal model hyperparameter combination. The target data is then input into the model to generate personalized rating results for each film and television content. The film and television content is then ranked according to the rating results to obtain a personalized film and television content recommendation sequence that meets the target user.

[0085] Step S5: Intelligent recommendation of film and television content. Based on the personalized film and television content recommendation sequence, a final list of recommended film and television content is generated and displayed on the user's end, realizing accurate matching and intelligent push of film and television content between users.

[0086] By performing the above operations, this solution addresses the technical problems of traditional film and television content recommendation methods, which rely solely on a single dimension of group or individual analysis, leading to homogenized recommendation results, difficulty in reflecting group differences, and insufficient adaptation to users' dynamic interests. These issues negatively impact the accuracy of film and television content recommendations and user experience. This solution innovatively combines group recommendation with personalized optimization. At the group level, users are divided into multiple groups based on clustering modeling, generating a recommendation set that reflects group characteristics. At the individual level, the group recommendation results are optimized and ranked by combining user personality and contextual data, achieving a dynamic balance between group representativeness and individual differences. This effectively overcomes the shortcomings of traditional recommendation methods in terms of the disconnect between group and individual modeling, improves the diversity and personalization of recommendation results, and ultimately achieves a comprehensive improvement in user satisfaction, activity level, and overall accuracy of film and television content recommendations.

[0087] Example 2, see Figure 1 and Figure 2This embodiment is based on the above embodiment. In step S1, the multi-source data acquisition is used to obtain the basic data required for film and television content recommendation. Specifically, it involves collecting data through the film and television content platform to obtain the raw data for film and television recommendation. The raw data for film and television recommendation includes historical film and television recommendation data, real-time film and television recommendation data, reference personalized ranking data, and target personalized ranking data. Both the historical film and television recommendation data and the real-time film and television recommendation data include user basic information data, user behavior data, and user preference data. The reference personalized ranking data and the target personalized ranking data include user basic information data, user behavior data, film and television content data, and recommendation environment data. The historical film and television recommendation data also includes historical recommendation feedback satisfaction. The reference personalized ranking data also includes reference... User feedback results; the film and television content data includes release date, country of origin, language version, film type, overall click-through rate, playback rate, completion rate, film duration, update status, and age rating; the user basic information data includes gender, age, region, registration time, and device type; the user behavior data includes click history, playback history, collection history, comment history, sharing history, search history, viewing duration, interaction rate, interaction duration, and viewing completion rate; the user preference data includes preferred video type, preferred actors, preferred video style, and active viewing time; the recommendation environment data includes the current time period, network status, and weather environment; the historical recommendation feedback satisfaction level has five levels: excellent, high, medium, low, and poor.

[0088] Example 3, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S2, the data optimization is used to optimize the collected raw film and television recommendation data. Specifically, it involves data cleaning, data standardization, data encoding processing, and film and television recommendation feature selection to obtain optimized film and television recommendation data. This includes the following steps:

[0089] Step S21: Data cleaning, used to process noise, missing items and duplicates in the original collected data, specifically to fill in missing values, remove outliers and process duplicates in the original data;

[0090] The missing value filling is used to handle missing fields in the original data. Specifically, it fills in missing numerical fields in the original data using the mean filling method, and uses the mode filling method for categorical fields.

[0091] The outlier removal is used to identify and remove data points that are obviously inconsistent with common sense. Specifically, it calculates the degree of deviation of each numerical field using the Z-Score algorithm and removes data samples that exceed a set threshold.

[0092] The duplicate item processing is used to remove records that are completely duplicated or have duplicate fields in the data. Specifically, it combines fields such as user and behavior timestamp for deduplication and uses duplicate row detection to remove duplicates.

[0093] Step S22: Data format conversion, used to convert heterogeneous data fields into a standard structured format that can be used for modeling and analysis, specifically time field conversion, text field conversion, and behavioral sequence data conversion;

[0094] The time field transformation is used to convert raw timestamp data into a time vector with structured meaning; specifically, it converts timestamp data into a structured time vector.

[0095] The text field conversion is used to normalize the original text fields, specifically by performing word segmentation, noise reduction, and vectorization on the text to generate semantic text vectors.

[0096] The behavior sequence data transformation specifically involves converting the behavior sequence into a fixed-length behavior feature vector.

[0097] Step S23: Data encoding processing, used to numerically encode discrete or categorical variables. Specifically, the one-hot encoding method is used to encode the discrete and categorical variables in the original data, perform numerical processing, and convert them into sparse binary vector form.

[0098] Step S24: Feature selection for film and television recommendations. This step involves selecting features that influence the film and television recommendation results. Specifically, it uses correlation coefficient analysis to extract feature variables related to film and television content recommendations from the original data, forming a set of recommendation features.

[0099] Example 4, see Figure 1 , Figure 2 , Figure 3 and Figure 4 This embodiment is based on the above embodiment. In step S3, the group film and television content recommendation is used to perform recommendation modeling on different types of user groups based on the clustering results of user groups, so as to generate a set of film and television content recommendations suitable for group characteristics. Specifically, it includes the following steps:

[0100] Step S31: Construct a clustering algorithm for user group segmentation, used in film and television content recommendation scenarios to divide a large number of users into several user groups based on multi-dimensional feature information, and generate corresponding user group segmentation models; specifically including the following steps:

[0101] Step S311: Similarity calculation, used to measure the differences between different users in the movie and TV recommendation scenario, specifically using Euclidean distance to calculate the similarity between data points; the data points are the data values ​​of each parameter for each user in the historical movie and TV recommendation data and the real-time movie and TV recommendation data, and the formula used is as follows:

[0102] ;

[0103] In the formula, This represents the similarity value between data point i and data point j. The vector representing the i-th data point. The vector representing the j-th data point. express Norm;

[0104] Step S312: Corrected density calculation, used to balance the differences between dense and sparse user groups in user group segmentation; specifically, by introducing a density correction factor to calculate the variation density, the corrected density value is obtained; the formula used is as follows:

[0105] ;

[0106] In the formula, This represents the maximum local density. This represents the corrected density value for data point i. Indicates the cutoff distance threshold. This represents the density correction factor, with a range of values. This is used to adjust the weights of sparse populations;

[0107] Step S313: Weighted distance calculation, used to enhance the ability to identify sparse groups and avoid the distance metric being biased only towards high-density groups. Specifically, when calculating the minimum distance to higher density points, dynamic density weights are introduced for correction to obtain the weighted distance value; the formula used is as follows:

[0108] ;

[0109] ;

[0110] In the formula, This represents the minimum distance from data point i to data points with a higher density. This represents the weighted distance of data point i. This represents the corrected density value for data point j;

[0111] Step S314: Cluster center formation, used in the user group segmentation process to screen and form the final cluster centers based on the comprehensive index of users' modified density and weighted distance. Specifically, it first selects a set of candidate representative points based on the dual conditions of modified density and weighted distance. Then, the candidate representative points are merged using the representative point merging threshold to finally generate a stable set of cluster centers.

[0112] The merging operation specifically involves calculating the similarity between any two data points in the candidate representative point set. If the similarity is... If these two representative points are too close and belong to the same user group, they should be merged, and the candidate representative point with the higher modified density should be retained as the new cluster center. If the cluster centers are different, then they are retained as different cluster centers; the formula used is as follows:

[0113] ;

[0114] ;

[0115] In the formula, Denotes the set of candidate representative points. This represents the average corrected density for all data points. This represents the threshold for merging data points, where n represents the number of data points. This represents the adjustment factor, used to control the threshold size;

[0116] Step S315: Initial micro-cluster construction, used to construct initial micro-clusters for user groups based on the established cluster centers, ensuring that representative users in high-density areas are correctly prioritized and assigned. Specifically, outliers are detected in the data points using the local anomaly factor method. Normal points that are close to the cluster center and are not abnormal are identified as candidate assignment objects, and these normal points are directly assigned to the group with the same nearest cluster center, thus forming local initial micro-clusters. The formula used is as follows:

[0117] ;

[0118] In the formula, Represents the set of normal points. Represents a local anomaly factor function;

[0119] Step S316: Residual point reallocation, used to reallocate the remaining user data points that could not be directly assigned to the initial micro-cluster, to ensure that boundary points and sparse points can be reasonably classified and to ensure the integrity of the overall group division; specifically including membership degree calculation, optimal cluster selection and residual point iterative allocation;

[0120] Step S3161: Membership degree calculation, specifically, for each unassigned data point, calculate the membership degree value of each data point to each cluster based on the dynamic fuzzy weighting method; the formula used is as follows:

[0121] ;

[0122] ;

[0123] ;

[0124] In the formula, This represents the similarity weight between data point i and data point j. This represents the nearest neighbor criterion weight between data point i and data point j. This represents the similarity weight between data point e and data point j. This represents the membership value of data point i to cluster c. This represents the set of k nearest neighbors of data point i. This represents the set of k nearest neighbors of data point j. This indicates the cluster to which data point j belongs;

[0125] Step S3162: Optimal cluster selection, specifically, for each unassigned data point, record its maximum membership value and the corresponding cluster; the formula used is as follows:

[0126] ;

[0127] ;

[0128] In the formula, This represents the maximum membership value of data point i. The optimal cluster to which data point i belongs;

[0129] Step S3163: Iterative allocation of remaining points, specifically, selecting the data point p with the highest membership degree among all unallocated data points and assigning it to the corresponding cluster. In, and update its allocation status, This indicates that the data point has been assigned. Simultaneously, for the k nearest neighbor data points q of data point p, their membership degrees to each cluster are updated, and the new maximum membership value and corresponding cluster label are recalculated. This assignment process is repeated until all remaining points are assigned. The formula used is as follows:

[0130] ;

[0131] In the formula, This represents the membership value of data point q to cluster c. This represents the nearest neighbor criterion weight between data point q and data point p. This represents the similarity weight between data point q and data point p;

[0132] Step S317: Obtain clustering results. Specifically, after all data points have been assigned, they are aggregated into the corresponding cluster labels to form multiple independent clusters, each cluster corresponding to a user group with similar characteristics.

[0133] Step S32: Obtain real-time user clustering results. Specifically, input the historical film and television recommendation data and the real-time film and television recommendation data in the film and television recommendation optimization data into the clustering algorithm for user group division, generate multiple independent user clusters, evaluate the clustering quality of each cluster, calculate its corresponding silhouette coefficient, and filter out the clusters with silhouette coefficients higher than a preset threshold to obtain the user group division results.

[0134] Step S33: Obtain the set of recommended film and television content for the target user. Specifically, based on the user group segmentation results, determine the group to which the target user belongs, statistically analyze and quantify the historical recommendation feedback satisfaction of the film and television content within the group, filter out the film and television content with excellent and high feedback levels within the group, and combine them to form the set of recommended film and television content for the target user.

[0135] By performing the above operations, this solution addresses the technical problem that existing clustering algorithms applicable to user group segmentation tend to favor high-density clusters when handling datasets with significant density differences, leading to the neglect or misclassification of sparse user groups, resulting in unstable clustering results and inaccurate group identification, thus affecting the accuracy of film and television content recommendation. This solution innovatively introduces a density correction mechanism, weighted distance calculation, representative point merging mechanism, and fuzzy weighted allocation strategy to improve the clustering algorithm. This effectively avoids the omission of sparse groups, ensures the accuracy of cluster center selection, and reduces the interference of outliers and noisy data on the clustering results, thereby improving the overall stability and reliability of clustering. It can more accurately characterize the user group structure, significantly improving the accuracy and diversity of film and television content recommendation, and enhancing the accuracy of group-based film and television content recommendation in complex user data environments.

[0136] Example 5, see Figure 1 , Figure 2 , Figure 5 and Figure 6 This embodiment is based on the above embodiment. In step S4, the personalized recommendation optimization ranking is used to perform personalized user rating and ranking on the group recommended film and television content set; specifically, it includes the following steps:

[0137] Step S41: Construct and train a personalized film and television ranking model to give personalized ratings to different film and television content and target users based on the candidate film and television content set recommended by the group; specifically, a personalized film and television ranking model is built based on a deep neural network, and reference personalized ranking data is used as the training data of the model to train the model. During the training process, mean squared error is used as the loss function, and the model parameters are jointly optimized through the backpropagation algorithm. The loss function value is continuously monitored and optimized until it converges, and the trained personalized film and television ranking model is obtained.

[0138] Step S42: Model hyperparameter optimization, used to improve the prediction accuracy of the personalized film and television ranking model. Specifically, it involves using an improved optimization algorithm to optimize the hyperparameters in the model to obtain the optimal combination of model hyperparameters; this includes the following steps:

[0139] Step S421: Initialize the search population individuals. Specifically, through random initialization, generate the position vectors of N search individuals in the parameter space. Each individual encodes a candidate model hyperparameter combination to form the initial search population.

[0140] The model hyperparameter combination includes learning rate, batch size, number of network layers, and number of hidden units;

[0141] Step S422: Calculate the individual fitness value, specifically by calculating the fitness value of the search individual in the population; and use the performance of the trained personalized film and television ranking model based on the search individual's position as the individual's fitness value.

[0142] Step S423: Adaptive reference individual selection, used to dynamically select the location of the interpolation point based on the individual's fitness value and its distribution in the search space, specifically for each search individual. Two random search individuals and The three individuals are ranked according to their fitness values ​​to determine the best individual. Name the remaining two individuals and Based on the positional relationship of the three individuals in the search space, the position of the adaptive reference individual is calculated. The formula used is as follows:

[0143] ;

[0144] In the formula, This represents a positive search factor, used to explore small areas on either side of the current individual. This represents a reverse search factor, used to extend the reference individual's location outwards;

[0145] Step S424: Perform a global exploration, specifically if Then, a global exploration operation is performed, using the adaptive reference individual position to guide the current individual's position update; the formula used is as follows:

[0146] ;

[0147] ;

[0148] In the formula, Represents the global exploration weight in the t-th iteration. This represents the scaling factor, with a value range of [0.2, 0.8]. , , and Both represent random numbers uniformly distributed within the range [0,1], and t represents the current iteration number. Indicates the maximum number of iterations. Indicates a random search for individuals. This represents the rounding function. Represents the natural logarithm function. This represents the position of the in-th search individual in the t-th iteration;

[0149] Step S425: Perform local exploration, specifically if If so, a local exploration operation is performed, and the position of the current iteration individual is updated; the formula used is as follows:

[0150] ;

[0151] In the formula, Indicates the globally optimal position. Represents a random number uniformly distributed in the range [0,1]. and These represent the upper and lower bounds of the search space, respectively. Represents the average function;

[0152] Step S426: Obtaining the optimal position of an individual. Specifically, after each iteration, the fitness values ​​of all individuals in the current population are evaluated. If the fitness of an individual's position is better than that of the current global optimal individual position, then the global optimal individual position is updated using that individual.

[0153] Step S427: Search iteration terminates, specifically when the fitness value of the search individual is higher than the fitness threshold and the maximum number of iterations is reached, the search is terminated and the global optimal individual position is obtained. The global optimal individual position specifically refers to the optimal combination of model hyperparameters of the personalized film and television ranking model.

[0154] Step S43: Obtain the optimal personalized film and television ranking model. Specifically, based on the optimal model hyperparameter combination, adjust the hyperparameters of the trained personalized film and television ranking model to obtain the optimal personalized film and television ranking model.

[0155] Step S44: Personalized film and television content recommendation and ranking, specifically, inputting the target personalized ranking data into the optimal personalized film and television ranking model, generating personalized rating results for each film and television content in the group recommended film and television content set, and ranking the film and television content in the group recommended film and television content set according to the rating results, thereby obtaining a personalized film and television content recommendation sequence that meets the target user.

[0156] By performing the above operations, this solution addresses the technical problems of unreasonable model parameter settings in existing personalized film and television ranking models, which lead to unstable and inaccurate personalized rating results for film and television content. It employs an adaptive reference individual selection strategy and an iterative global exploration weight improvement optimization algorithm to perform a global and local joint search on key parameters in the personalized film and television ranking model. This automatically obtains the optimal combination of model hyperparameters, effectively improving the intelligence and accuracy of hyperparameter search, significantly enhancing the accuracy and stability of the model output results, and achieving stability in personalized film and television content recommendation and ranking optimization.

[0157] Example 6, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S5, the intelligent recommendation of film and television content specifically involves generating a final list of recommended film and television content based on the personalized film and television content recommendation sequence, and displaying it on the target user's film and television recommendation interface, thereby achieving accurate matching and intelligent recommendation between the user and film and television content.

[0158] Example 7, see Figure 1 and Figure 2 This embodiment is based on the above embodiments, and the technical solution adopted by the present invention is as follows: The intelligent recommendation of film and television content based on artificial intelligence provided by the present invention includes a multi-source data acquisition module, a data optimization module, a group film and television content recommendation module, a personalized recommendation optimization and sorting module, and an intelligent recommendation module for film and television content;

[0159] The multi-source data acquisition module obtains raw film and television recommendation data through data acquisition operations and sends the raw film and television recommendation data to the data optimization module;

[0160] The data optimization module receives data sent by the multi-source data acquisition module, performs data cleaning, data standardization, data encoding processing and film and television recommendation feature selection on the raw film and television recommendation data to obtain optimized film and television recommendation data, and sends the optimized film and television recommendation data to the group film and television content recommendation module and the personalized recommendation optimization ranking module.

[0161] The group film and television content recommendation module receives data sent by the data optimization module. Specifically, it constructs a clustering algorithm for user group segmentation based on an improved clustering algorithm, inputs the film and television recommendation data in the film and television recommendation optimization data into the clustering algorithm, generates user group segmentation results, combines the historical recommendation feedback information of each group, filters and combines the film and television content within the group, obtains the group recommended film and television content set, and sends the data to the personalized recommendation optimization and sorting module.

[0162] The personalized recommendation optimization and ranking module receives data sent by the data optimization module and the personalized recommendation optimization and ranking module. Specifically, it constructs and trains a personalized film and television ranking model, and obtains the optimal model hyperparameter combination by combining the improved optimization algorithm to obtain the optimal personalized film and television ranking model. The target data is input into the optimal personalized film and television ranking model to generate personalized rating results for each film and television content. The film and television content is ranked according to the rating results to obtain a personalized film and television content recommendation sequence that meets the target user. The data is then sent to the film and television content intelligent recommendation module.

[0163] The intelligent film and television content recommendation module receives data from the personalized recommendation optimization and sorting module, generates a final film and television content recommendation list based on the personalized film and television content recommendation sequence, and displays it on the user's end, realizing accurate matching and intelligent push of film and television content between users.

[0164] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0165] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0166] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. An intelligent recommendation method for film and television content based on artificial intelligence, characterized by: The method includes the following steps: Step S1: Multi-source data collection. Through data collection operations, the raw data for film and television recommendations is obtained. Step S2: Data optimization, specifically involves cleaning, standardizing, encoding, and selecting movie and TV recommendation features from the original movie and TV recommendation data to obtain optimized movie and TV recommendation data; Step S3: Group-based film and television content recommendation, used for recommending film and television content at the user group level. Specifically, it involves improving the clustering algorithm by introducing a density correction mechanism, weighted distance calculation, representative point merging mechanism, and fuzzy weighted allocation strategy to construct a clustering algorithm for user group segmentation. Film and television recommendation data from the optimized film and television recommendation data is input into the clustering algorithm to generate user group segmentation results. Combined with historical recommendation feedback information for each group, film and television content within the group is filtered and combined to obtain a set of recommended film and television content for the group. This includes the following steps: Step S31: Construct a clustering algorithm for user group segmentation; specifically including the following steps: Step S311: Similarity calculation, specifically using Euclidean distance to calculate the similarity between data points; Step S312: Corrected density calculation, specifically, involves calculating the variability density by introducing a density correction factor to obtain the corrected density value; the formula used is as follows: ； In the formula, This represents the maximum local density. This represents the corrected density value for data point i. Indicates the cutoff distance threshold. This represents the density correction factor, with a value range of [value range missing]. Used to adjust the weights of sparse populations. This represents the similarity value between data point i and data point j; Step S313: Weighted distance calculation. Specifically, when calculating the minimum distance to higher density points, dynamic density weights are introduced for correction to obtain the weighted distance value; the formula used is as follows: ； In the formula, This represents the minimum distance from data point i to data points with a higher density. This represents the weighted distance to data point i; Step S314: Cluster center formation, specifically, firstly, a set of candidate representative points is selected based on both modified density and weighted distance conditions. Then, the candidate representative points are merged using the representative point merging threshold to finally generate a stable set of cluster centers. The merging operation specifically involves calculating the similarity between any two data points in the candidate representative point set. If the similarity is... If these two representative points are too close and belong to the same user group, they should be merged, and the candidate representative point with the higher modified density should be retained as the new cluster center. If the cluster centers are different, then they are retained as different cluster centers; the formula used is as follows: ； In the formula, This represents the threshold for merging data points, where n represents the number of data points. Indicates the adjustment factor; Step S315: Initial micro-cluster construction, specifically, involves detecting outliers in the data points using the local anomaly factor method, identifying normal points that are close to the cluster center and are not anomalous as candidate assignment objects, and directly assigning normal points to the group with the same nearest cluster center, thereby forming local initial micro-clusters; the formula used is as follows: ； In the formula, Represents the set of normal points. Represents the local anomaly factor function. The vector representing the i-th data point; Step S316: Residual points redistribution; specifically including the following steps: Step S3161: Membership degree calculation, specifically, for each unassigned data point, calculate the membership degree value of each data point to each cluster based on the dynamic fuzzy weighting method; the formula used is as follows: ； ； In the formula, This represents the similarity weight between data point i and data point j. This represents the nearest neighbor criterion weight between data point i and data point j. This represents the similarity weight between data point e and data point j. This represents the membership value of data point i to cluster c. This represents the set of k nearest neighbors of data point i. This represents the set of k nearest neighbors of data point j. This indicates the cluster to which data point j belongs; Step S3162: Optimal cluster selection, specifically for each unassigned data point, recording its maximum membership value and the corresponding cluster; Step S3163: Iterative allocation of remaining points. Specifically, select the data point p with the highest membership degree among all unallocated data points, allocate it to the corresponding cluster, and update its allocation status. At the same time, update the membership degree of the k nearest neighbor data points q of data point p to each cluster, and recalculate the new maximum membership degree value and the corresponding cluster label. Repeat the above allocation process until all remaining points are allocated. Step S317: Obtain clustering results. Specifically, after all data points have been assigned, they are aggregated into the corresponding cluster labels to form independent clusters. Each cluster corresponds to a user group with similar characteristics. Step S32: Obtain real-time user clustering results. Specifically, input the historical film and television recommendation data and the real-time film and television recommendation data in the film and television recommendation optimization data into the clustering algorithm for user group division, generate multiple independent user clusters, evaluate the clustering quality of each cluster, calculate its corresponding silhouette coefficient, and filter out the clusters with silhouette coefficients higher than a preset threshold to obtain the user group division results. Step S33: Obtain the set of recommended film and television content for the group. Specifically, based on the user group segmentation results, determine the group to which the target user belongs, statistically analyze and quantify the historical recommendation feedback satisfaction of the film and television content in the group, filter out the film and television content with excellent and high feedback levels in the group, and combine them to form the set of recommended film and television content for the target user. Step S4: Personalized recommendation optimization and ranking, used to optimize the group recommendation ranking of film and television content at the individual user level. Specifically, it involves building and training a personalized film and television ranking model, and improving the optimization algorithm by adopting an adaptive reference individual selection strategy and iterative global exploration weights. The model hyperparameters are optimized to obtain the optimal model hyperparameter combination, resulting in the optimal personalized film and television ranking model. The target data is input into this model to generate personalized rating results for each film and television content, and the film and television content is ranked according to the rating results to obtain a personalized film and television content recommendation sequence that meets the target user. Step S5: Intelligent recommendation of film and television content. Based on the personalized film and television content recommendation sequence, a final film and television content recommendation list is generated and displayed on the user's end, realizing intelligent push of film and television content to users.

2. The intelligent recommendation method for film and television content based on artificial intelligence according to claim 1, characterized in that: In step S4, the personalized recommendation optimization ranking specifically includes the following steps: Step S41: Construct and train a personalized film and television ranking model. Specifically, a personalized film and television ranking model is established based on a deep neural network. Reference personalized ranking data is used as the training data for the model. The model is trained to obtain the trained personalized film and television ranking model. Step S42: Model hyperparameter optimization; Step S43: Obtain the optimal personalized film and television ranking model. Specifically, based on the optimal model hyperparameter combination, adjust the hyperparameters of the trained personalized film and television ranking model to obtain the optimal personalized film and television ranking model. Step S44: Personalized film and television content recommendation and ranking. Specifically, the target personalized ranking data is input into the optimal personalized film and television ranking model to generate personalized rating results for each film and television content in the group recommended film and television content set. The film and television content in the group recommended film and television content set is ranked according to the rating results to obtain a personalized film and television content recommendation sequence that meets the target user.

3. The intelligent recommendation method for film and television content based on artificial intelligence according to claim 2, characterized in that: In step S42, the model hyperparameter optimization specifically includes the following steps: Step S421: Initialize the search population individuals. Specifically, through random initialization, generate the position vectors of N search individuals in the parameter space. Each individual encodes a candidate model hyperparameter combination to form the initial search population. Step S422: Calculate the individual fitness value, specifically by calculating the fitness value of the search individual in the population; and use the performance of the trained personalized film and television ranking model based on the search individual's position as the individual's fitness value. Step S423: Adaptive reference individual selection, used to dynamically select the location of the interpolation point based on the individual's fitness value and its distribution in the search space, specifically for each search individual. Two random search individuals and The three individuals are ranked according to their fitness values ​​to determine the best individual. Name the remaining two individuals and Based on the positional relationship of the three individuals in the search space, the position of the adaptive reference individual is calculated. The formula used is as follows: ； In the formula, Indicates a positive search factor. Indicates the reverse search factor; Step S424: Perform a global exploration, specifically if Then, a global exploration operation is performed, using the adaptive reference individual position to guide the current individual's position update; the formula used is as follows: ； ； In the formula, Represents the global exploration weight in the t-th iteration. This represents the scaling factor, with a value range of [0.2, 0.8]. , , and Both represent random numbers uniformly distributed within the range [0,1], and t represents the current iteration number. Indicates the maximum number of iterations. Indicates a random search for individuals. This represents the rounding function. Represents the natural logarithm function. This represents the position of the in-th search individual in the t-th iteration; Step S425: Perform local exploration, specifically if If the current optimal individual position is used, a local exploration operation is performed to guide the position update of the current iteration individual. Step S426: Obtaining the optimal position of an individual. Specifically, after each iteration, the fitness values ​​of all individuals in the current population are evaluated. If there is an individual whose fitness is better than the current global optimal individual position, then the global optimal individual position is updated using that individual. Step S427: Search iteration terminates. Specifically, when the fitness value of the search individual is higher than the fitness threshold and the maximum number of iterations is reached, the search is terminated and the globally optimal individual position is obtained. The globally optimal individual position specifically refers to the optimal combination of hyperparameters of the personalized film and television ranking model.

4. The intelligent recommendation method for film and television content based on artificial intelligence according to claim 1, characterized in that: In step S5, the intelligent recommendation of film and television content specifically involves generating a final list of recommended film and television content based on the personalized film and television content recommendation sequence, and displaying it on the target user's film and television recommendation interface, thereby achieving accurate matching and intelligent recommendation between the user and the film and television content.

5. The intelligent recommendation method for film and television content based on artificial intelligence according to claim 1, characterized in that: In step S1, the multi-source data collection specifically involves collecting data through a film and television content platform to obtain raw film and television recommendation data. The raw film and television recommendation data includes historical film and television recommendation data, real-time film and television recommendation data, reference personalized ranking data, and target personalized ranking data. Both the historical film and television recommendation data and the real-time film and television recommendation data include user basic information data, user behavior data, and user preference data. The reference personalized ranking data and the target personalized ranking data include user basic information data, user behavior data, film and television content data, and recommendation environment data.

6. The intelligent recommendation method for film and television content based on artificial intelligence according to claim 1, characterized in that: In step S2, the data optimization specifically includes the following steps: Step S21: Data cleaning, specifically, filling in missing values, removing outliers, and processing duplicates in the original data; Step S22: Data format conversion, specifically time field conversion, text field conversion, and behavioral sequence data conversion; Step S23: Data encoding processing, specifically, using one-hot encoding to encode the discrete and categorical variables in the original data, performing numerical processing, and converting them into sparse binary vector form; Step S24: Selecting film and television recommendation features. Specifically, the correlation coefficient analysis method is used to extract feature variables related to film and television content recommendations from the original data to form a set of recommendation features.

7. An AI-based intelligent recommendation system for film and television content, used to implement the AI-based intelligent recommendation method for film and television content as described in any one of claims 1-6, characterized in that: It includes a multi-source data acquisition module, a data optimization module, a group film and television content recommendation module, a personalized recommendation optimization and sorting module, and a film and television content intelligent recommendation module; The multi-source data acquisition module obtains raw film and television recommendation data through data acquisition operations and sends the raw film and television recommendation data to the data optimization module; The data optimization module receives data sent by the multi-source data acquisition module, performs data cleaning, data standardization, data encoding processing and film and television recommendation feature selection on the raw film and television recommendation data to obtain optimized film and television recommendation data, and sends the optimized film and television recommendation data to the group film and television content recommendation module and the personalized recommendation optimization ranking module. The group film and television content recommendation module receives data sent by the data optimization module. Specifically, it constructs a clustering algorithm for user group segmentation based on an improved clustering algorithm, inputs the film and television recommendation data in the film and television recommendation optimization data into the clustering algorithm, generates user group segmentation results, combines the historical recommendation feedback information of each group, filters and combines the film and television content within the group, obtains the group recommended film and television content set, and sends the data to the personalized recommendation optimization and sorting module. The personalized recommendation optimization and ranking module receives data sent by the data optimization module and the personalized recommendation optimization and ranking module. Specifically, it constructs and trains a personalized film and television ranking model, and obtains the optimal model hyperparameter combination by combining the improved optimization algorithm to obtain the optimal personalized film and television ranking model. The target data is input into it to generate personalized rating results for each film and television content. The film and television content is ranked according to the rating results to obtain a personalized film and television content recommendation sequence that meets the target user. The data is then sent to the film and television content intelligent recommendation module. The intelligent film and television content recommendation module receives data from the personalized recommendation optimization and sorting module, generates a final film and television content recommendation list based on the personalized film and television content recommendation sequence, and displays it on the user's end, realizing accurate matching and intelligent push of film and television content between users.

Images