Research On Personalized Recommendation Algorithmbased On Multi Attribute Ratings

Today in the 21st century,the era of information explosion,everyone in the face of information has hundreds of millions of the,especially in the electronic commerce website how users where find their interested information,has not only limited to the users themselves to search for,need more E-commerce Recommender systems to help users find the information he is interested in.So the research field of recommender system has become more and more important,it can recommend the most interested information for the user.At present,most of the recommender system is the user of the product evaluation information personalization recommendation,the explicit or implicit evaluation information is expressed as a preference level of users to be graded projects in a single dimension,the single dimension score information can not be effective expression and the difference of degree of preference for users of a product all aspects,and affect the recommended algorithm is recommended for the performance.In view of the traditional single dimension score of the recommendation algorithm based on,multi attribute rating based Recommender Systems Consider users of all aspects of the production information to evaluate differences of personalization recommendation decisions.The main work of this paper includes the following aspects:First,in consideration of the multiple attributes of the hotel rating based on,improved based on multi attribute scores of the collaborative filtering recommendation algorithm of the three methods of empirical analysis and comparison of their accuracy and diversity.The three methods are,based on multi attribute similar average collaborative filtering,based on multidimensional distance collaborative filtering algorithm and multi attributes of AHP based collaborative filtering algorithm.Secondly,through the analysis of user rating attributes on the preference of the hotel,the weight of the linear programming model is introduced to find the attribute in the past,this paper proposes a collaborative filtering algorithm based on multiple attribute linear programming.Again,how to measure the indicators of quality of a recommendation system.There are many,this paper mainly discusses the popular accuracy and diversity,although the recommendation accuracy is undoubtedly important,but scholars increasingly recognize accuracy does not always mean that are useful to the user,perhaps the user wants to provide the goods with diversity.Therefore,in addition to the analysis accuracy,this paper also consider another important index of diversity recommendation quality,the relationship between accuracy and diversity is discussed.Finally,in order to validate the proposed method and empirical,we collect the real data on the hotel website.Collected 165829 users on the hotel's multi attributes score(respectively,cost-effective score,comfort score,location score,health score,sleep score,service score).Mean absolute error(MAE)and diversity are used to measure the performance of the algorithm.The experimental results show that the proposed method can significantly improve the recommendation accuracy and user diversity in multi attribute environment.