Research On Purchase Decision And Personalized Recommendation Of Retail Commodity

Commercial activity is a very important part of state economy and is very important in national economy.With the development of science and technology and the popularity of the internet,business practices are becoming more information and more intelligent.A large amount of data has been accumulated in the commodity trading business especially retail sales.It becomes a very hot problem how to analyze the data to get valuable information and use the information to improve commerce.However,because of the different data structure and data missing problems,it is difficult to data mining for business data.Based on the above background,this thesis conducts on a series of research work about the application of data mining technology in the retail data.Specifically,after analyzing the research and applied status of data mining in retail data,this thesis aims at the features of business data and commercial activities and proposes a solution of data mining in retail data.The main work and results obtained from the researches of this thesis are summarized as follows.1,Based DCCB network,this thesis proposes an algorithm for personalized recommendation.The algorithm transfers the relationship between customer and goods to the relationship between goods and goods.Using the goods as nodes and relationships between goods as edges,a DCCB network is built.The weight of edges is updated as time goes by.In order to reduce difficulty of calculation,the nodes and edges which have low weight are deleted to downscale the network.Finally,the algorithm based on improved modularity is used to recommend goods.The experiment compared this algorithm with other recommendation algorithm on same data set,proving that the recommendation algorithm based on complex network is more effective and more efficient than collaborative filtering recommendation algorithm.2,This thesis proposes an algorithm for commodity purchase decision.The attributes of price of purchase are got using method named SWMN.The goods similar with target goods are found by analyzing the description of target good and attributes of good's price.The historical prices of similar goods are used to complete data of target goods and build the training data.The decision model is created by decision tree algorithm using training data,and is used to decide whether to buy or not to buy goods.According to experiments on real data,it proves that the algorithm is effective.3,In order to combine theory and application,this thesis designs and implements a consumer-oriented commercial data mining system.The system is based on architecture.Web crawler technology is used to get data from the internet.The recommendation algorithm and commodity purchase decision algorithm which are purchased in this thesis are implemented in this system.The users of the system can get personalized goods recommendation server and commodity purchase decision service.The results of these services are showed using visualization technology.