Research On Multidimensional Dynamic Recommendation Technology For Mobile E-Commerce Platform

In the background that the amount of data and information on the Internet are is increasing exponentially,a good recommendation system has become a guide for people in the information ocean,as it can study and analyze the personalized information such as user's behaviors,interests,habits and preferences to construct user's knowledge model,interest model and other models that can deduce user's characteristics,and to provide customers with the information that accords with user's characteristics.At present,most of the existing recommendation algorithms at home and abroad are based on single-dimensional and static,either on time-based collaborative oversight,or only considering the "user-location" two-dimensional.Such recommendation algorithms are increasingly unsuitable for mobile environments with multi-dimensional and dynamic features.For example,the location-based collaborative filtering algorithm mainly obtains the "user-location" score matrix by analyzing the location information trajectory of mobile users,and constructs the user preference model based on location information.The recommendation algorithm based on time collaborative filtering,mainly through the analysis of the characteristics of mobile users' interest changes in time,takes into account the forgetting rule,interest persistence,attenuation,popularity and seasonality to model,which has possessed certain timeliness and dynamic characteristics.The main contribution of this thesis is to study the recommendation technology from one-dimensional and static ones to multi-dimensional and dynamic ones,to propose a multi-dimensional dynamic collaborative filtering recommendation algorithm which integrates multi-factors such as location,project and time,and to apply it to the current mobile e-commerce platform.Experimental results indicate that the proposed algorithm outperforms the chief one-dimensional and static algorithms in the aspect of accuracy.