Research On Heat Conduction Recommendation Algorithm Based On The Bipartite Network

With the advent and development of the era of big data,information on the Internet showing explosive growth.It is becoming more and more difficult to get the information we want from the vast amounts of information quickly and accurately,this is the problem of information overload.In order to solve the problem,recommendation system arises at the historic moment.The core of the recommendation system is the recommendation algorithm.There are many kinds of recommendation algorithms,one of which is the combination of the graph model and the heat conduction in physics.We call it the heat conduction recommendation algorithm based on the bipartite network.Because of its novel combination and high diversity of the recommendation results,the heat conduction recommendation algorithm has been widely concerned.However,the accuracy of the proposed algorithm is poor,this paper analyzes the causes of this result,and puts forward the corresponding improvements,the main work is as follows:(1)In this paper,an ununiformed resource allocation heat conduction recommendation algorithm based on the bipartite network is proposed.First of all,in the heat conduction algorithm based on the bipartite network,we need to initialize resource value of all items.The initial resource value of each item representatives the recommendation power of this item.In the original algorithm,the initial resources of the items with a score greater than or equal to the threshold are set to 1,ignoring the item's own personality.In order to solve this problem,we introduce a tuning parameter to change the initial resource allocation,and do experiments on different data sets.The experimental results show that the improved algorithm can effectively improve the accuracy of the proposed algorithm.(2)This paper proposes a hybrid algorithm which combines the recommendation algorithm based on the item similarity and heat conduction recommendation algorithm.The hybrid algorithm can reduce the influence of data sparsity on the recommendation results.As we all know,when the density of the dataset is dense,the more information is contained in the dataset,the more accurate of the recommendation result is.On the contrary,if the dataset is sparse,the available information is relatively less,the accuracy of the recommendation result is bound to be affected.Therefore,we improve the density of datasets to improve the accuracy of the results.First of all,we need to find the highest scored item of the target user,and then calculate the similarity between the highest scored item and the target user's unmarked items based on cosine similarity,then we choose some items with high similarity and regard this items as the target user's marked items.On this basis,we recommend items for the target user applying the heat conduction recommendation algorithm based on the bipartite network.(3)In this paper,we propose an algorithm combining association rules and heat conduction recommendation algorithm.This algorithm intends to reduce the influence of data sparsity on the recommendation results.Firstly,the algorithm find the nearest neighbors set of the target user's highest scored item by using association rule mining algorithm,then regard this items as the target user's marked items to fill the data set and increase the density of data set.On this basis,we recommend items for the target user applying the heat conduction recommendation algorithm based on the bipartite network.