Remission Data Sparsity Question In Collaborative Filtering

With the rapid development of computer and Internet technology,mankind has entered the information society and the Internet age,people enjoy the Internet brings convenience,but also difficult to quickly and accurately locate the information they need from the mass of information.This is the "information overload" problem.In this context,the recommendation system was born.Collaborative filtering recommendation algorithm is today the most widely used system of an algorithm.But there are simple and efficient,while data sparsity,cold start,scalability and other issues.Data sparsity makes use collaborative filtering recommendation algorithm system faces a serious shortage.Aiming at this problem from the perspective of their own features two category experts and projects of users,ease of data sparseness problem,so as to improve the recommendation accuracy.Firstly,the data filling algorithm based on expert weights category,the algorithm from the perspective of project categories to improve the resolution of the data,allowing users to focus point increase from a single product to the catalog.Then use the algorithm to select the category of expert and user data to be filled similarity as weights by category of project experts predict user ratings to be filled on the project rating,a rating matrix dense degree rise.Finally,experimental verification of the forecast filling algorithm can improve the recommendation accuracy.Secondly,due to the traditional collaborative filtering algorithms start from the user's score matrix,resulting in a sparse matrix,the emergence of similarity calculation is not accurate,affect the outcome of the recommendation.Proposes a collaborative filtering algorithm step,first according to the project characteristics enable filling of sparse data users can have more neighbors users then join in the traditional demographic similarity calculation formula to improve the similarity similarity between users accuracy.Thereby improving the quality of recommendation.Finally,use MovieLens data sets were designed in accordance with Section A total of three sets of experiments to verify the proposed method,experiments show that the mean filling more than the traditional method and the K-means clustering method based on the category of experts filling scoring matrix method reliability,lower MAE value.The user step by step collaborative filtering method can collaborative filtering algorithm in the case of sparse data than traditional users to provide better recommendation results.