Research On Recommendation Algorithm Based On Improved Random Forest

With the development of social economy,e-commerce has become an indispensable part of life.The exponential growth of information in e-commerce makes it difficult for users to quickly and accurately find the goods of interest in the mass of commodity information.The personalized recommendation algorithm is created in this context.Recommendation algorithm changes the way of e-commerce receiving requests from passive users to actively recommending them.It also solves the shortcuts for users to find their favorite items from the information overload network.In this paper,we use improved RandomForest algorithm into recommendation algorithm.Random Forest algorithm is statistics theory that combins the set of decision tree classification and it has feature subspaces to construct the modelcan deal with noise and avoid over fitting surpassingly.In this paper we mainly introduced the several classic methods of Random Forest algorithm and their characteristics.Researching algorithms in domestic and overseas were analyzed and summarized systematically from the process of the construction of the decision forest,and we propose an improved method for random forest algorithm.In this paper,we propose an improved stochastic forest algorithm with support vector machine and stochastic forest algorithm.The basic weak classifier is the decision tree in the random forest,and the decision tree is the most powerful one in choosing the classification ability.In this paper,combining the support vector machine(SVM)algorithm with attribute selection of decision trees,the hyperplane of linear combination(support vector)of feature variables is divided,which is more powerful than single attribute classification.It has been improved in the process of random forest decision tree construction.The experimental results show that the improved random forest algorithm has high accuracy.In this paper,we use the real user behavioral data of Alibaba to establish the improved random forest algorithm model by mining user behavior,and finally get a list of recommended items for users.Experiments show that it can effectively forecast and recommend the future purchase of items,which is of great significance to the development of recommendation algorithm under the premise of user's historical behavior data.