Research On Collaborative Filtering With Temporal Dynamics And The Application

With the rise of electronic commerce, more and more people choose to use it go shopping, also request higher on the shopping experience. Personalized recommendation system is to increase the customer shopping experience, widely used in reality. The most successful is the application of personalized recommendation algorithms is collaborative filtering recommendation algorithm.This paper analyzes the shortcomings of the existing collaborative filtering algorithm, and the research status on the time dynamics, points to an important question-- commodities have successively purchase order. The traditional collaborative filtering algorithms only consider the degree of correlation between users or goods, and not according to the sequential pattern mining and recommendation of goods. This thesis combines the advantages of the existing algorithms, joined the time dynamic characteristic, proposed a collaborative filtering time considering dynamic improvementalgorithm recommended comprehensive, a nd tested by electronic commerce real sales data, verify its recommendation effect is better.The main contributions of this paper include the following:(1) Through a review of research on the existing literature, find the direction of the current research focus and research of scholars, summing up the characteristics of the existing research, analyzes the shortcoming of the traditional collaborative filtering recommendation algorithm, sum up the algorithm only mostly in a one or two aspects of improvement, research and join time dynamic properties also only stay in the user the interest of the change, which leads to the direction and the focus of this study.(2) Combining with the excellent improvement direction in current research and improvement method, effect of adding screening goods sequence pattern and the removal of hot effect, put forward a comprehensive collaborative filtering recommendation algorithm solution. Its core idea can be summarized as the first of the original data set preprocessing, including to noise and clustering; and then computing the nearest neighbor with a new similarity measure function, the similarity calculation function considered hot coefficient, score factor and time factor; finally use the GSP algorithm for sequential pattern mining commodities, on the recommendation results set filter again, to solve the user to buy after the mouse to the recommended computer problems.(3) A collection of real data verify the Amazon reviews, experimental design and recommend on the effect of the improved algorithm proposed in this paper. Compared with the traditional collaborative filtering algorithm, improves the accuracy of recommendation, and improved the data sparseness problem existing in the traditional algorithm, verified the effect of the improved algorithm is more excellent recommendation.