Study On Recommendation System Algorithm Based On Web Data Mining

With the rapid development of Internet, there is more and more information on theInternet, and the number of information is exponential growth. How to find valuableinformation and knowledge in these vast amounts of information is one of the mostimportant problems for every e-commerce businesses. As a comprehensive technologywhich involves many areas, Web data mining aims to extract hidden useful knowledgeinformation from massive, irregular, noisy data. Only the results of Web data miningapplied in practical item can ultimately benefit users, which involves an importantapplication of Web data mining—recommendation.Recommendation systems based on Web data mining results, firstly analyze theusers`behavior preferences, and then recommend to the users something they needed.The core of recommendation system is algorithm. This paper detailed analysis and studyon association rules and clustering algorithm of recommendation algorithms, andimproved the Apriori algorithm and the K-means algorithm. Association rules andClustering are two popular methods in recommendation system. Apriori algorithm is aclassical algorithm in the association rule algorithm. This paper improved the two stepsof rules generated. The concept of set array was introduced in the step of frequent itemsgenerated, and the concept of tree was introduced in the step of rules generated. Theclustering result of K-means clustering depends on the initial clustering center, thispaper gave two kinds of new methods to select initial clustering center. The last, thispaper analyzed the application of association rule and clustering algorithm inrecommendation system. Two kinds of algorithm combination could improve theprecision of recommend, and improve the performance of recommendation system.The improved Apriori algorithm could reduce the times of database scanned andavoid the redundant rules generated. The new methods optimized the way for selectinginitial clustering center, which improved the accuracy of clustering.