A Study On Collaborative Filtering Algorithm Similarity And Parallel Implementation

With the Internet into the Internet era,the network data production rate has undergone tremendous changes.The recommendation system is able to know the changes of users ’needs and the changes of users’ interest in real time.It realizes the content that users are interested in,and the recommendation system has become a solution of large data information overload.The recommendation algorithm analyzes the similarity between users and projects by using the behavior data of the users on the Internet,and recommends the items that may be of interest to the users by analyzing the similarity relationship.As the scoring matrix of users is very sparse,the traditional collaborative filtering algorithm has the problem of low prediction accuracy and poor scalability,which leads to the large deviation between the similarity and the actual situation.In this paper,we propose a new user-based predictive algorithm,named New Item-based Collaborative Filtering,to provide a high-accuracy recommendation for the problem of unreasonable candidate sets in traditional recommendation algorithms.Item-based CF algorithm first adds the association rules to the similarity algorithm,improves the traditional Pearson similarity calculation method to calculate the similarity relation among the items,and then constructs the inter-item correlation matrix,and then uses the association matrix to predict The user scoring matrix for the item.In this paper,the average absolute error and root mean square error are used to evaluate the prediction results.The experimental results show that the NItem-based CF algorithm improves the prediction accuracy.In order to solve the problem of inefficient operation of network environment,this paper uses Spark distributed computing platform to parallelize the implementation of NItem-based CF algorithm.The experimental results show that the efficiency of the algorithm is improved after the parallelization.In addition,the ALS algorithm is implemented on the Spark platform in order to solve the problem of the large dimension of the recommendation data set.The ALS is realized by the alternating least squares method.The experimental results show that the ALS algorithm alleviates the problem of high dimensionality of the data to a certain extent.