Research On Recommendation Algorithm Based On Granularity Sequence Pattern

The recommendation system is a mass of information filtering system,which could be used to predict the user's rating or preferences.The common methods of recommendation include collaborative filtering recommendation,content recommendation,knowledge recommendation,association rules recommendation,social trust recommendation and hybrid technology recommendation.Among them,the content-based recommendation algorithm has a wide range of applications,and the implementation method is simple,which can overcome the data sparse,cold start,new items and other referral system common difficulties successfully.In the research of content-based recommendation algorithm,this paper proposes a recommendation algorithm based on granularity sequence pattern,aiming at the problem of low correlation and insufficient feature extraction in complex projects.Firstly,analyzing the relationship between project properties and then prioritizing the importance of project properties.The feature matrix of project is obtained by granularing and calculating the contribution degree of each particle size after sorting.Secondly,the user's granularity is generated according to the user's historical behavior information,and the user preference matrix is extracted by Apriori algorithm.Finally,the product operation of the project feature matrix and user preference matrix is made,and the results can predict the preference probability in the improved sigmoid function,thus the Top-N project recommendations are completed.The input of experiments is the MovieLens dataset,and the results show that the accuracy of recommendation algorithm based on granularity sequence pattern is higher than that of state-of-the-art recommender algorithms.It is also demonstrated that our algorithm achieved is than collaborative filtering recommendation algorithms when the number of users is the same.The overall performance of the algorithm is better than other recommended algorithms.