Integrating Similarity And Machine Learning Into Weighted Slope One Algorithm

Because of its high prediction accuracy and being not limited by project type, collaborative filtering recommendation algorithm has been widely concerned. This thesis makes a deep research in Slope One algorithm, which is a classic algorithm in collaborative filtering recommendation algorithm based on memory, it has attracted much attention due to advantages such as simple calculation efficiency and so on. But when calculating, the algorithm is not considering the similarity between users or items, and calculation process is taking too much space in memory, moreover, the advantage of prediction accuracy is not obvious than traditional collaborative filtering algorithm. Therefore, this article puts forward some improvement measures as follows:First, the paper has introducing the knowledge of collaborative filtering, including the research of background and current situation, etc. Then, the research foundation and related knowledge of Slope One algorithm are introduced in this paper, including their advantages and disadvantages and the improvement measures.Secondly, since the Slope One algorithm in computing are not considering the problem of similarity, so, based on the original algorithm, we proposed a new algorithm, called Integrating User and Item Similarity into Weighted Slope One Algorithm. Meanwhile, we use the trust mechanism and Jaccard coefficient weighted method to calculate the similarity between the users respectively, and select the neighbor users, then, using the Pearson correlation coefficient calculation method to calculate the similarity between items, in the end, the combination of two hybrid recommendation algorithms is proposed.Then, considering the memory based method is simple and intuitive, are easily to understand, but the method based on model is generally get more accurate prediction results and running faster than the method based on memory, so this paper combining the advantages of them to improve the Slope One algorithm, and put forward the Integrating Machine Learning into Weighted Slope One Algorithm, and using the least squares method in machine learning to improve Slope One algorithm.Finally, we verified the proposed algorithms through the experiment. These algorithms are all conducted in contrast experiment of two different datasets, and give them different analysis of the results. Results show that the improved algorithm in this paper compared with the original Slope One algorithm, greatly improve the accuracy of prediction.