Research On Key Technologies Of Movie Recommendation Based On Spark

With the rapid growth of Internet video traffic,streaming media services such as online movies have become an important part of people's demand for network services.As the quality and quantity of online movies continue to increase,users' attention to online movies has gradually increased.How to push from the huge online movie resources according to user needs is an important issue to improve user satisfaction and optimize network resources.The online movie recommendation system is one of the effective ways of film promotion technology,and has become a research hotspot.In recent years,the film recommendation system has been mainly based on the user's rating of the film to find relevant users,but different users have different standards for reviewing the movie,which will result in different evaluations of the same movie,and even a big difference.Happening.Therefore,there are certain limitations in recommending movies based solely on the results of the ratings.In order to improve the objectivity of film evaluation,film reviews are included in the evaluation criteria.The analysis and search for relevant users through the emotional information embodied in the user's movie reviews has become the research focus of the current film recommendation key technologies.At present,most of the film recommendation techniques based on sentiment analysis divide emotion into three categories: neutral,positive and negative.In fact,the user's emotional expression on the movie is complex.There are three types of information to judge.There is a big limitation;In technology,there are also problems such as insufficient semantic understanding ability and weak strength analysis,resulting in low accuracy of sentiment classification.In view of the above research shortcomings,this thesis comprehensively analyzes the emotional categories,strengths,and semantic factors such as negative words and degree words,establishes the rules for emotional evaluation of movie film reviews,and expands the comments emotions into “good” and “music”.Seven kinds of emotions such as "anger","evil","sorrow","fear" and "shock".At the same time,an improved strategy for finding similar users is proposed.Firstly,the sentiment analysis is performed on the text,and the Top-k similar users of the target users are searched by emotional similarity.The similar users are used to fill the sparse matrix,and based on this,the user features are combined.Similarity and traditional score-based collaborative filtering similarity to further find users with more interesting interests,and finally obtain user similarity and generate recommendation results through linear weighting calculation.The experimental results show that the proposed film recommendation technique has certain advantages compared with several classical methods.Finally,based on the proposed recommendation strategy,this thesis designs and implements a Spark-based film recommendation system.The system test results show that the system can provide personalized recommendations according to the needs of different users.