Research On User Data Mining And Behavior Analysis Of O2O Project

With the rapid development of the Internet,the national social economy to a new level.At the same time,Internet technology has actually changed people's daily life.The rapid development of e-commerce makes the traditional industry is facing a crisis,and O2O(Online To Offline)model for most traditional industries to find novel business opportunities.O2O model is an extension of the traditional e-commerce model,the core value of this model is to start from the user's real needs,to provide users with high quality service experience.It is not only simple to the user from online to offline or offline to online,but also to the maximum user-friendly.Through analysis of user behavior data and dig out the valuable information,on the one hand can provide recommendation service for users,help the enterprise to obtain the higher viscosity users,on the other hand can help enterprises for precise marketing.At present,there are still relatively few researches on user data mining in O2 O mode,and the accuracy of existing personalized recommendation techniques need to be further improved.In this paper,discover hidden valuable information available to the enterprise by mining large amount of user behavior data,to help business provide accurate service push.The main work of this paper is as follows:(1)Research on the low accuracy of collaborative filtering recommendation algorithm,considering the influence of number of users to score the overlapping project on the similarity calculation,the consistency of the user score overlapping project into the similarity calculation,calculation method of similarity is further modified,alleviate the negative impact of excessive estimation,in order to improve the accuracy of recommendation algorithm.(2)Clustering algorithm and collaborative filtering algorithm characteristics were discussed,aiming at the massive user data,clustering processing by user attributes,narrowing the search range similar neighbor users,reducing the computational overhead,improve the recommendation performance.(3)For new users,users can be clustered according to the attributes of the cluster to find similar users,and then recommend to solve the cold start problem.(4)Through two standard data sets,the cross validation method is used to evaluate the algorithm.The experimental results show that the proposed collaborative filtering recommendation algorithm based on user clustering can achieve good accuracy and accuracy,and it is helpful to improve the quality of recommendation.In addition,consider the user behavior sequence in the enterprise application,the frequency statistics of user behavior data,get a probability score matrix as theapproximate score,solve the implicit feedback behavior cannot be directly used for collaborative filtering algorithm problems,but also alleviate the data sparsity.The wangpupos for SAAS Intelligent cash register user behavior data in the clouds for experiment and results show that the strategy has better recommendation effect.