Information Recommendation Based On Web Data Mining

This paper systematically expounded Web data mining algorithms and its entire mining process in real Web environment, then this paper researched and analyzed related Web data mining mining algorithms involved in the personalized recommendation system. Algorithm include similarity algorithm, collaborative filtering algorithms and so on.For the shortcomings of the traditional and improved collaborative filtering algorithm, this paper raised a based on user recessive evaluation Combo collaborative filtering algorithm, the core of the algorithm is first of all to improve the traditional dominant scoring mechanism, and integration the user’s implicit rating information into user-dominant scoring mechanism, and for non-rated items using a simple default value processing, then for modified item score datas using the hierarchical clustering analysis to obtain a hierarchical clustering tree, then combined with relevant domain knowledge to find abstract knowledge from a hierarchical clustering tree, and to build a hierarchical tree kinds of users evaluation model and mapping function F which from clustering tree to hierarchical tree. Finally, when recommended some info for target user, Combo algorithm will first analyze item info which the target user has scored, and then finded items types that item related through the mapping function F, and from user-types score set finding the "nearest neighbors", then produced accurate recommendation.Finally, this paper designed a simplified recommendation system framework and done Combo algorithm experimental simulation, and this paper prove that this algorithm in the face of score datas sparseness problems and network Pushing Hands cheating problems encountered in personalized recommendation system showed a very good recommendation result.