Design And Implementation Of Business Data Preprocessing And Association Mining Modules In Keyword Optimization System

Electronic Commerce is becoming a popular way of shopping,and more and more business dealers publish their products online for customers to inquire and purchase.During this process,electronic commerce websites set information for their products(such as product name,description and product keyword)and publish them online,while potential buyer search interesting products through the interface provided by electronic commerce website.However,as electronic commerce websites and customers have different understanding and various descriptions of the products,the product keyword set by business dealer may not match the search keyword used by potential buyer,which makes it hard for the product to be easily accessed by target customer.And this will directly influence the service that the electronic commerce websites provide for their clients.The thesis introduces the design and implementation of business data preprocessing and association mining modules of the keyword optimization system which is used by an electronic commerce business company.These modules find the consumers'search pattern through analyzing the business data(such as access logs,inquiry logs and product categories),and give company keyword optimize recommendations in order to improve closing rate.The main works in this thesis are as follows:(1)Preprocess data using weighted trajectory in order to solve the problem that the size of the access logs are too large which largely reduces the efficiency of data mining procedure.(2)Use weighted Apriori algorithm to find association rules in weighted trajectory and discover consumers' search pattern.(3)Use incremental update algorithm to generate association rules when weighted trajectory is updated,thus increase the mining efficiency.Based on these aspects,the business data preprocessing and association mining modules which include the function of building mining data source,mining search pattern based on association rules and mining incremental data based on dynamic threshold are designed and tested.The tests show that these modules guarantee the performance while getting effective results.