Collaborative Filtering Recommendation Algorithm Based On Hellinger Distance Similarity Measure

In recent years,with the rapid development of the Internet,people have entered the era of big data,all kinds of information data are achieving explosive exponential growth,and the amount of data information is in an overload state.As an effective means to solve the problem of information overload,personalized recommendation system has been widely used in various fields,such as e-commerce platform,movie recommendation,music recommendation and so on.The continuous optimization and improvement of recommendation algorithm can not only improve the user experience,but also bring certain commercial value and promote the development of other industries.Recommendation algorithm is the most important part in recommendation system.Good recommendation algorithm can predict user behavior quickly and accurately.Among so many personalized recommendation algorithms,collaborative filtering algorithm,because of its simple,efficient and accurate characteristics,stands out from many personalized recommendation algorithms and becomes the most widely used and classic recommendation technology.Because of the increasing of data,the collaborative filtering algorithm is faced with data sparsity and cold start of system.Domestic surgical researchers have also proposed their own solutions to this problem.By avoiding or alleviating the above problems,we can effectively improve the accuracy of the recommendation results.This dissertation proposes a novel similarity measurement method,which is a recommendation algorithm based on Hellinger distance similarity.This method is different from the similarity measurement method commonly used in collaborative filtering algorithm.It calculates the similarity between items from the perspective of scoring probability distribution.In order to verify the effectiveness of this method,three comparative experiments are designed.First of all,the results of the traditional similarity calculation method are compared with the results of the similarity measurement based on Hellinger distance.The experimental results show that the similarity measurement method used in collaborative filtering recommendation algorithm is feasible,and the algorithm has good accuracy.Secondly,this paper proposes a method of classifying and granulating the data set,and calculates the item similarity in the classified data set.Through the comparison experiment with unclassified data,it is verified that the similarity measure after classification can improve the accuracy of the algorithm.Finally,the paper compares the proposed algorithm with slope one algorithm to explore the overall performance difference between the proposed algorithm and the classic collaborative filtering recommendation algorithm.The method proposed in this paper can effectively avoid the data sparsity problem of the scoring matrix in the process of similarity calculation,and the similarity calculation results are more accurate.