Medical Recommend Algorithms Based On Collaborative Filtering

With the rapid development of internet technology and the arrival of the information age,the amount of data presents an explosive growth trend.Since the users have to face large amount of redundant and complicated data,there are many obstacles to easily and conveniently obtaining the information that they to find.Thus,as a promising topic,recommendation technology has emerged and developed gradually.As the core of the recommendation technology,the recommendation algorithm has been widely used in various fields.But the proposed algorithm based on the collaborative filtering still has three problems,which are spars data,cold boot and extensibility.Therefore,in view of the above problems,this paper used the medical system as the background and the medical data to do the research and carried out the following work.Since the non-registered patients do not have the authority to comment,the system cannot obtain the user rating matrix of unregistered patients.This work proposes the concept of preference score matrix for the click and collection data of unregistered patients,then a medical recommend algorithm based on non-registered patients is proposed.The algorithm first defines the feature attributes of target objects.Secondly,according to the different users’ behaviors,the algorithm considers each user’s preference for different target objects in each behavior.Then the similarity degree of users should be calculated based the similarity degree calculation formula of the different behaviors,so as to implement the clustering operation.Finally,combined with the users set with similar interests and the rating information of other users to predict the score of unrated objects and obtains the recommendation results.For registered patients,the traditional approaches often rely on the user rating matrix for recommendation.However,not all patients will evaluate the doctors or only a few registered patients will evaluate doctors.Therefore,there is a case where the user rating matrix is empty or sparse.Based on the clicks,collections and rating data,this work proposes the concept of user preference attribute matrix,and combines with user rating matrix to propose a medical recommendation algorithm for registered patients.The algorithm firstly defines the user score matrix and the user preference property matrix.Then,the similarity of the two matrices is calculated,and the weight of each matrix is given to comprehensively measure the similarity between users.Finally,the algorithm predicts the score of unrated objects and obtains the recommendation results.For the two algorithms,this work compares and analyzes the existing collaborative filtering algorithms respectively,and proves that the two proposed algorithms are more accurate than the existing algorithms,which makes the recommended results more consistent with the user’s preferences.