Research On Collaborative Filtering And Application Of Recommendation System In The Medical Field

Search engine technology is becoming the main ways and methods when user filtering information, at the same time, the recommendation technology based on search engine came into being, by predicting user’s interests to recommend the most interested in or the most needed information for the user. In terms of referring users to the target object, the most widely used algorithm is collaborative filtering algorithm. At present, collaborative filtering recommendation algorithm mainly exist three major technical problems:cold start, data sparseness and big data calculation.To solve the three problems, this paper studied and designed the recommendation algorithm in medical field, using the hadoop cloud platform for data distributed storage, solve the mass data storage problems. Recommend service system introduced hierarchical analysis model to evaluate target object, then use collaborative filtering technology based on user interest clustering, first of all, analyze the characteristics of the target object attribute, establish tags and classified the object, and use the analytic hierarchy method for object attribute, establish comparative matrix, statically evaluate target object, so as to solve the problem of cold start; Secondly, using the similarity between object classes, and relation between user and the target object, using k-means method to cluster the user from the user’s behavior logs within one cluster, mining user interest preference, computing similarity between users, choose the the highest score of nearest neighbor users, recommended list is created for the target user, to reduce data sparseness.this paper verified recommend accuracy in two ways:first, recommend accuracy improved by the collaborative filtering recommendation algorithm based on users interest this paper puts forward compared with the traditional collaborative filtering technology;Second,compared doctors and hospitals data in the service recommendation system with recommend data in the medical website, and changes of user behavior data traffic, this recommend system have higher accuracy and higher degrees.