The Study And Application Of Recommendation Technology Based On Pattern Recognition

In the information explosion day, recommendation system which is an effective organization form of information is receiving more and more attention. Pattern recognition is introduced into Recommendation system in this paper to classify the users, and provide recommendation service for users according to distribution characteristics of user's interest in the classification results.First of all, the basic concept, the research progress and research methods of recommendation system and pattern recognition are introduced, which provides a theoretical basis for this article.Secondly, recommendation algorithm based on pattern recognition is designed in the paper. The ratings of the project are resolved into the rating of the label of Interest based on MovieLens data which is set as an example. Paper sets up interest preference vectors for users according to user's interest in different labels mean score distribution. Users are classified by the characters of user's interest which come from interest preference vectors of users.Third, this paper proposes a feature selection of users based on the K-means clustering method.Paper constructs feature subset according to the user's interest characteristic collection.K-means clustering algorithm is used to determine the best classification number for each feature subset and DBIndex is used for feature selection as judgment function.Correlation characteristics are deleted and retained features is requested.Fourth, in order to search the similar users more accurately and quickly, paper puts forward the concept of characteristics groups of user on the basis of users by the interest groups, which use smaller feature subset to represent the whole group's interest preferences.Finally, on the basis of pattern recognition and the characteristics of user groups, paper designs a collaborative filtering recommendation algorithm based on the user.