Research And Implementation Of Movie Recommendation System Based On LorentzFM

In recent years,China’s Internet industry has benefited from the growth of endogenous demand.Under the digitizing tide,the network penetrates into all aspects of human social life.With the explosive growth of Internet data,the problem of information overload has become more and more obvious.How to help users filter information efficiently and reasonably is not only a satisfaction of users’ high-quality needs,but also a long-lasting way for enterprises to stay at the core of the industry.For this reason,the recommendation system comes into being,which provides users with personalized recommendation services by exploring the relationship between users and information,and helps users quickly locate content related to their interests.This paper is based on the research of recommendation algorithm,puts forward a combination model which considers both memory and generalization ability,and designs and implements a personalized movie recommendation system.First,this paper introduces the recommendation process of the recommendation system and related technical theoretical basis,analyses the current mainstream shallow and deep interaction models,and discusses their advantages and disadvantages.To preserve the ability of shallow models to learn and utilize features that occur frequently in historical interactions,while retaining the ability of deep models to convey their relevance to higher-order features,and to mine the correlation between sparse features and final tags,this paper establishes the infrastructure of composite models.Then,the feature engineering of the current CTR algorithm model is briefly introduced.For the current mainstream feature interaction methods such as inner product,outer product,Hadama product,bilinear and DCN ideas,which do not take into account the geometric significance of feature interaction,hyperbolic space is selected as the feature embedded representation space.The fine structure between the features and the special properties of the triangular inequality can be constructed by the Lorenzian distance between two points under the Lorenzian metric.This paper presents a Lo Fm&Deep model based on the above and compares it with several mainstream models in three open datasets to obtain the best results.Experiments show that the model has better characterization ability and better classification effect.Finally,this paper designs and implements a personalized movie recommendation system based on the proposed recommendation algorithm.The system is developed based on B/S architecture and consists of front-end and back-end,database and recommendation system.In this paper,aiming at the unfriendly situation of the traditional multi-recall method to the user’s personalized recommendation,the recall module and rough sorting module are combined into a simple unified recall to enhance the user’s personalized recall needs.In addition,the system includes many functions such as popular recommendation,network popularity recommendation,movie collection and movie retrieval,which can meet the user’s needs for personalized movie recommendation.