A Co Clustering Personalized Recommendation Research Based On Grey Relation

With the development of 4G technology, Mobile Internet has tremendous potential for development, the global access to a new era of mobile Internet, Chinese have begun to enter the era of mobile Internet, convenience and real-time are the biggest characteristics of mobile Internet, but faced with the age of big data, the convenience of mobile Internet bas being threatened, how to find effective data in big data has become a bottleneck restricting to the development of the mobile Internet. Personalized recommendation based on the historical data of user to analysis user’s interest, so as to realize the personalized information recommendation. In the background of big data and mobile Internet, at present how to combine data mining, knowledge discovery and other technology to achieve more effective personalized recommendation has become a research hotspot. Although people have pay attention to the recommendation system from proposed, but the problem of sparsity, scalability, real-time, accuracy and cold start are still the bottleneck of its development.The paper arms to the research of personalized recommendation technology, At first, this paper details the research status on personalized recommending, and then put forward the matrix filling algorithm based on the attributes and improved similarity calculation algorithm, to achieve a more accurate user similarity measure, In order to improve the real-time performance and scalability of the recommendation system, the paper puts forward a recommendation algorithm based on weighted clustering, the specific contents are as follows:First of all, the paper advances a matrix filling technology based on their attributes, calculating the user’s initial score according to the characteristics of users and items, and using the linear weighted method to get the final initial score, and filling the user rating matrix, which reduces the sparsity of the matrix. At the same time, using the linear weighting method to improved the cosine similarity calculation algorithm, and proposes a improved similarity calculation method based on it.Furthermore, designs a recommendation model based on C-clustering, the model including two stages of offline and online, at offline,using the feature to complete filling the initial score matrix,according to the weighted clustering algorithm to cluster user and item, and according to the improved similarity calculation method, to calculate the similarity. At online stage,according to the users submiting score, to determine the category of the user and project, in the corresponding class using the similarity weighted pattern of project and user to generate k nearest neighbors, and implemented Top-N recommendation.Finally, aiming at the scalability problem, the paper puts forward a similarity incremental updating model. based on the the initial category judgment of target user,, avoid the huge amount of calculation at all update set. At the same time, the use of independent factors, for each additional score, as the original similarity model and independent factor, reduce the similarity calculation. In order to improve the real-time performance of the system, make up the scalability problem of recommendation system. The results show that, Co clustering Personalized Recommendation Research based on grey relation by combining the technology of features、gray correlation analysis、matrix filling、clustering,effectively alleviats the problem of collaborative filtering, achieves more accurate, real-time recommendation, and enhances the expansibility of the system.