Design And Implementation Of TV Programrecommendation System Based Onrandom Walk Model

With the vigorous development of network and information technology,network data has shown explosive growth.Although there are a wide variety of data information around people,there are not many data information that real users are interested in.Therefore,how to mine users’preferences from massive and complex data information,so as to recommend products that users may be interested in,is the core issue of the recommendation system.In recent years,smart TV has gradually entered thousands of households and become one of the most important forms of entertainment in family life.There are a large number of network videos on smart TV for users to choose,which makes it difficult for users to select and determine the programs they are interested in from a large number of program sources.At the same time,program providers are also trapped in providing suitable program sources to interested users,resulting in difficulties in content-based marketing.Therefore,the establishment of a TV program recommendation system that can meet the needs of many smart TV users has broad application prospects.Program recommendation algorithms usually reflect users’preferences based on users’ direct ratings of programs.However,TV program operators generally do not set up the function of users’ ratings of TV on demand programs.Even if they can score,users need to score through the remote control.Due to the cumbersome operation,it is difficult to encourage users to spend more time scoring,resulting in the poor availability of the current rating based recommendation system,A new program recommendation system based on user behavior is needed.By analyzing users’ historical viewing behavior and using valuable evaluation indicators to obtain users’ preferences,this paper obtains users’ scoring matrix for TV programs,and solves the problem that there is no explicit scoring for users in the current TV on demand program recommendation system.At the same time,in the application environment of the real recommendation system,due to the large number of users and programs,compared with the large number of users and programs,users’ program viewing records are very limited,resulting in a serious sparse scoring matrix.In order to alleviate the problem of sparse data,this paper introduces a random walk model to mine the indirect similarity between programs,and fills in the user program initial scoring matrix to improve the accuracy of recommendation.Aiming at the problem of new users’ cold start,this paper proposes to cluster users and reflect their preferences according to the preference information of similar users,so as to solve the problem of cold start.