Research On The Application Of Data Mining Technology In The Shop Product Recommendation

In the current e-commerce platform, there are tens of thousands of various types of branch stores, for each shop, only provide one kind of product is not enough, at least need to display a variety of goods to meet the various needs of multiple consumers, provide consumers with more choice. Recommendation system using e-commerce sites to provide customers goods and information to help them decide to buy what goods, recommend information generally located at the top of the site, including content have access to the consumer statistics, or through the analysis of the past of consumer purchase behavior to predict a new purchase intention. Recommend information include provide personality products for consumers; summarize other consumer opinions and comments.Data mining is aimed at a large number of data by using the method of mathematical analysis concluded that exist in the patterns and trends, and discovery of useful information. Data mining and Web mining from large amounts of data knowledge discovery process without human intervention is closely related. The source of data for web mining knowledge is the Internet. Information filtering system using web data mining technology in view of the two types of network: based on the content filtering system abstract knowledge from the web documents, and collaborative filtering system to use network user’s information.To solve the scalability and sparse data problem in collaborative filtering, this paper puts forward the recommendation system solution is to use the user clustering technology and project technology. User clustering and clustering technology work of similar evaluation is to identify the user groups and projects. Once you create a cluster, through the calculation of average opinion of cluster can predict target audience rating. In some clustering technology users will be scattered in a number of clustering, clustering algorithm can generate a fixed size partition, or based on some similarity threshold will be generated in conformity with the requirements of different size partition number. The results of the prediction for the whole cluster carried out in accordance with the participation of the weighted average. Combined with the user clustering and project collaborative filtering is much more scalable and more accurate than the traditional method.Recommendation system designed in this paper, first of all, collect information of users and projects, including personal information, user browse the history of commodity information, the user text for goods made by evaluation and rating information, e-commerce information basic attributes of goods, commodity trading records information and writing evaluation and rating information for the goods. For users and project information collected to do basic data preprocessing, and then use the user clustering and project do collaborative filtering algorithm, for not rating goods do forecast rating, and recommended for sending information in the form of web pages to the server, finally displayed on the user’s browser.System testing collected data set to choose nearly ten thousand data as the training data set, about nearly thousands of users to include about 2000 books, and form data in the database table. Predict behavior design of recommendation system of evaluation indicators for accuracy and average coverage. After data table, data preprocessing, parameter Settings, user clustering, project, comprehensive analysis and conclusion recommend testing and validation testing of multiple steps, proved in this paper, basic design of the electronic commerce recommendation system meet the need.To improve the practicability of recommendation system, finally, the paper designed the book recommendation system application platform, application platform is designed for the front desk and the background of two parts, including the front desk website functionality has registered user login, history inquiry, online recommendation, commodities ratings, reviews, etc. Background management includes recommended management, user management, books management, and the shopping cart. One of the key management functions for the recommended management, the other three functions of management of aided design. Administrator after every once in a while, through the library books management view no rating information, using the recommended management functions, first set the recommendation algorithm, and then set the user similarity threshold, such as project similarity threshold value and forecast not rating book to make recommendation on; Data management is of the collected books and user data preprocessing; Recommendation on management can view and manage for not rating books recommended by this article system to forecast rating.