Research On The Weighted Slope One Recommandation Technology Based On Clustering

The explosion of information scale in the Internet, meets the needs of the users of information. But the large amount of information makes it hard for users to locate the useful information quickly, reduces the utilization of information, leads to the emergency of information overload. Personalized recommendation technology is a kind of user-oriented effective means of personalized recommendation, which core is the recommendation algorithm.Slope One algorithm, a simple and effective collaborative filtering algorithm based on the project, can achieve good recommend results in the case of a small amount of data to achieve good results, which has been widely used. But the existed Slope One algorithm can’t make a precise recommendation under the condition of sparse data. And Slope One algorithm will utilize the irrelevant items to predict and unable to perceive the change of user interest quickly.In order to solve the above problem, in this paper the calculation method of weight is improved, and put forward the improved weighted Slope One algorithm, and then introduce the related technology of data mining, do data’s classification and pretreatment, the weighted Slope One algorithm based on clustering is proposed. The main work done is as follows:First, on the basis of the traditional K- Means algorithm, put forward a kind of automatic generation of clustering center of K- Means algorithm based on minimum spanning tree, effectively solve the traditional K- Means algorithm caused by the initial clustering center to select the randomness of the local optimal problem, improve the clustering effect;Second, based on the clustering results to the original project score matrix to predict filling, solve data sparseness’ s problem in algorithm, and according to the clustering results to reduce the recommended candidate set’s size, reduce recommendation algorithm computation;Third, considering project attributes and project evaluation of similarity degree of different project, introduce of the project properties and projects score of comprehensive similarity calculation method, improve the accuracy of project similarity;Fourth, in order to reflect the variation of the user’s interests in the algorithm, stand out new data’s function to weaken the old data. Add time weight in recommendation algorithm, considering the factors that influence the time weight, propose to join the visit frequency weighting function of time;Fifth, according to the improved algorithm is proposed in this paper, design recommendation system, introduce the composition of system module, call relations between modules and module’s internal algorithm process, using Movie Lens data set to validate on the system.Experiments show that the weighted Slope One algorithm based on clustering compared with the traditional recommendation algorithm, the addition of clustering algorithm can effectively solve the data sparseness, reduce amount of calculation; Project similarity and time weight to join improve the accuracy of the prediction algorithm and time sensitivity. Integral algorithm has obvious decrease on average absolute error, can effectively improve the overall performance of recommendation system.