A Research Of Personalized Recommendation Based On Comment Data Mining

With development of information technology,more and more data generated in the internet,along with information overload.To obtain significant information in mass data for users,personalized recommendation becomes a great choice.Personalized recommendation is capable of providing users the goods or information they may like.Usually the feedback from users would contain ratings and comment data.Developers are supposed to get to know what users like through a large amount of information in users' comments.By analyzing user comment data,developers are able to know users' preference better and thus improve precision of their algorithm.Up to now,personalized recommendation and related studies are making progress.Due to the limitation of rating information,traditional recommendation method which is based on rating information has some precision issues.Therefore,making good use of users' comment information would definitely help developers to do personalized recommendation.Around this issue,this paper proposes an algorithm to do recommendation based on the main points of users comment.Based on users' comment information,developers could model user preference and item property to show matching degree between these two and do enhanced treatment for the result of prediction of final rating.In this way,developers can effectively reduce interference factors and finally improve precision of recommendation.In addition,this paper proposes another recommendation algorithm which is a combination of Factorization Machine and Hidden Markov.This algorithm is able to produce great result even under the situation where the data is insufficient.As the contrast experiment can tell,two recommendation algorithms mentioned in this paper:one is based on comment data and the other is the combination of Factorization Machine and Hidden Markov,has the ability to effectively improve precision of recommendation.