Personalized Media Recommendation Algorithm Based On Popularity Prediction

Context-aware Recommendation System has become one of the most activeresearch fields in personalized recommendation. The main research task ofcontext-aware recommendation technology is how to make use of user contextinformation, and furtherly improve the accuracy of the recommendation results anduser satisfaction on the basis of traditional recommendation technology. Because ofits pervasive computing and personalized features, context-aware recommendationtechnology is widely used in many industrial fields. In recent years, CARS has broadapplication prospects especially in electronic commerce, information retrieval, mobileapps, movies and music to recommend, electronic tourism and other fields. Althoughresearch in context-aware recommendation technology has made some progress, but itis still an emerging research field full of problems and challenges. The problems ofcomplex context information types and different user needs, make it difficult torecommend the appropriate resources according to the user’s context information andtheir needs, thus, the user satisfaction degree is low. Therefore, on the basis ofcomplex user context and different requirements, how to reach the more accuraterecommendation results and higher user satisfaction has become a problem to besolved.Based on the above analysis and the digital home environment, therecommendation technologies related to context awareness are researched, then,aiming at the problems existing in the research environment and the shortages oftraditional recommendation technologies, popularity prediction technology wasimproved, and multiple linear regression techniques is adopted for forecasting thepopularity. This paper studies inadequate combination of the context and new interestpoint in CARS. On the basis of the traditional recommendation technology andaccording to the characteristics of the context information in digital home environment, a Personalized Media Recommendation Algorithm based on PopularityPrediction (MRAPP) is put forward. Firstly, The algorithm analyzes, classifies andmodels the user context information in digital home environment, a recommendationresult is calculated through the traditional Content-based Recommendationtechnology(CB), then, for content-based Recommendation technology is unable torecommend new interest point for the user, the concept of popularity is introduced.Finally, through theoretical proof and simulation experiment, the effectiveness ofMRAPP algorithm is demonstrates.The innovation points of this paper is considering the user is not satisfied withrecommendation results similar to context information, and trying to recommend newinterest points. Therefore, popularity prediction value are calculated by multivariatelinear regression on the server, and by combining results of content-basedrecommendation technology, new resources suitable to user’s environment isrecommended and the user satisfaction is improved.Researches of the personalized media recommendation algorithm have become ahot topic of recommendation technology. In the case of a large amount of information,an effective media recommendation algorithm should well combined with usercontext information and ensure the accuracy of resources recommended to the user.At the same time of guaranteeing the accuracy of resources, research onrecommending new media interest point for users has theoretical and practicalsignificance.