Design And Implementation Of Web-based Personalized Movie Recommender System

With the continuous growth of the scale of Internet users and the number of movie resources in my country,the movie system is facing a serious problem of "information overload",making it difficult for users to obtain effective movie resources.Search engines provide a solution to the problem of information overload,but when users do not have clear viewing goals,search engines are difficult to function.In order to solve the problem that it is difficult for search engines to meet the needs of customers who do not have a clear viewing target,this paper builds a web-based personalized movie recommendation system to provide users with personalized movie recommendation services and solve the problem that users are difficult to choose in the case of information overload.The main research contents are as follows:Firstly,the commonly used algorithm models in recommender systems are introduced.By analyzing the advantages and disadvantages of the existing algorithm models,the algorithm model used in this paper,the ensemble tree model,is determined.Then,the basic principles of ensemble learning and related technologies in the construction of web systems are introduced to provide necessary preparations for subsequent algorithm analysis and system construction.Secondly,in view of the low prediction accuracy of the existing ensemble tree model,a LightGradient Boosting Machine based on long and short term feature embedding is proposed.Solved the problem of the ensemble tree model.Aiming at the problem that high-dimensional discrete features are difficult to learn,a word embedding expression method Word2 Vec is proposed,which converts high-dimensional sparse vectors into low-dimensional dense vectors,which is beneficial to the learning of high-dimensional discrete features by ensemble tree model.Aiming at the problem of high cost of artificial feature combination,the process of leaf node splitting in the gradient boosting tree model and the relationship between the various levels of the tree model are used to simulate the process of feature selection and feature combination respectively,so as to model feature engineering and reduce the cost of manual feature engineering.cost.Aiming at the problem that users’ long-term and short-term interests may be biased,a feature embedding method based on long-term and short-term memory units is proposed,which effectively encodes users’ long-term and short-term interests through long-term memory units and short-term memory units.The similarity of long-term and short-term interests enables the ensemble tree model to better distinguish users ’ long-term and short-term interests and improve the recommendation quality of the model.Compared with the experimental results of the baseline model on the MovieLens-20 M and MovieLens-10 M datasets,compared with the LightGBM model before the improvement,the AUC indicator of LSTFE-LightGBM on the MovieLens-20 M dataset It increased by 1.7%,and the F1-Score indicator increased by 2.4%.On the MovieLens-10 M dataset,the AUC index is increased by 0.5%,and the F1-Score index is increased by 1.2%,which proves the effectiveness of the proposed model.Finally,through the analysis of the functional requirements and non-functional requirements of the personalized movie recommendation system,a personalized movie recommendation system based on the LSTFE-LightGBM recommendation algorithm is designed,and the functions of each part of the system are tested.