Unified Collaborative Filtering Model Based On Combination Of Latent Features

With the constant expanding of e-commerce and the rapid growth of commodity inthe number and types, customers have to spend a lot of time to find the goods they want.Much time is wasted on scanning irrelevant information and products. The informationoverload could make the e-commerce consumers less and less. In order to solve thisproblem, personalized recommendation system came into being. Personalizedrecommendation system is an advanced business intelligence platform based on massivedata mining technology. It can provide information service for e-commerce to helpcustomers make decisions. It can also be used in other kinds of information platformsuch as interest communities like Douban Movie to provide dynamic recommendationsof objects users may be interested in. some kinds of systems even can recommendcross-cutting, which mine users’ interests and preference more deeply. Personalizedrecommendation systems make people take more use of information platforms andincrease the loyalty of customers.This paper firstly gives a brief introduction of recommendation systems. And then,it studies these typical recommendation algorithms. The paper tries to find a moreeffective algorithm which can fix some shortcomings of traditional ones. CollaborativeFiltering (CF) is one kind of recommendation algorithms which is widely used in manycommercial systems. It has high recommending accuracy and good expansibility. CFcan be classified as memory-based and model-based ones. And memory-based CFcontains user-based CF and item-based CF, but both kinds have shortcomings of datasparsity and scalability. Model-based CF are developed using data mining, machinelearning algorithms to find patterns based on training data. These are used to makepredictions for real data. Most of the models are based on creating a classification orclustering technique to identify the user based on the test set. There are severaladvantages with this paradigm. It handles the sparsity better than memory based ones.This helps with scalability with large data sets. It improves the prediction performance.In this paper, we propose a model-based CF algorithm which is a unified methodcombining the latent and external features of users and items for accuraterecommendation. A mapping scheme for collaborative filtering problem to text analysisproblem is introduced, and the probabilistic latent semantic analysis was used tocalculate the latent features based on the historical rating data. The main advantages of this technique over standard memory-based methods are the higher accuracy, constanttime prediction, and an explicit and compact model representation. The experimentalevaluation shows that substantial improvements in accuracy over existing methods canbe obtained.