Personalized Recommendation Model And Application Based On Itemrank And User Preferences

With the rapid development of electronic commerce,there is a lot of information on the website.It provided a huge challenge and opportunity for customers and e-commerce companies.By the rapid development of network technology,the number of web pages is growing rapidly,and the users want to find the information that they needed,which spend a lot of time.Personalized recommendation technology can be based on user's historical behavior,mining users' interest,and then recommend the information or goods the users needed,and also reduce search time.In the increasingly competitive environment,the personalized recommendation system is not only a means of business marketing,but can enhance the users' adhesion,and increase user loyalty,and bring huge commercial interests to e-commerce.This paper introduces the main content of the collaborative filtering recommendation algorithm and the PageRank search algorithm.With collaborative filtering recommendation algorithm calculating the user similarity,does not consider traditional commodity weights.So combining the PageRank algorithm with the collaborative filtering algorithm,propose a kind of algorithm(LTR)that consider the commodity rankings and users' preferences.First,calculate the ranking of commodities based on PageRank algorithm,optimize the similarity function,and improve the recommendation accuracy.Divide into two stages: At the first,by setting the higher ranking of commodities a higher weight,calculate the user similarity according to the weight;At the second stage,based on user similarity to find a neighbor,recommend the high weight products which bought by the neighbor to the user,as a result of improving the quality of precision.Secondly,add the score coefficients into the similarity function.With considering the influence of the coefficient of the score,calculate the similarity of users.The similarity function is optimized,and then the user similarity is calculated.Finally,by using the Movielens database 100,000,make a verification of the data.Do a total of seven experiments to verify that improved algorithm is better than the previous recommendation algorithm.