| Summary: | 碩士 === 大同大學 === 資訊工程研究所 === 90 === Our approach aims to establish a useful inference model that allows enterprises to recommend most appropriate items to customers to buy in the near future once the customers have some purchasing records in our database or web server. Many data mining models provide efficient methods to discover the embedded relationships between items sold to customers. However, the association rules from database only reflect whether interesting items appear frequent enough in the “old” transaction records. Not many traditional models illustrate how to apply the available attributes of associated items to fulfilling the personalized service in the e-commerce. Based on the available transaction records, we use AprioriTid model to derive the association rules from large database. The dug associated items as well as useful attributes associated with items, such as division number and sale price, are then become the crucial factors of training patterns for the back propagation model. The clustering results from ART-2 model and the transforming methodology from clustered results to the neural network are discussed in our data mining model. How to derive the association rules from large database, how to apply the derived rules to training a neural network for inference purpose, and how to implement a personalized service system are illustrated by simple examples. Simulation results from different models are summarized in the tables to verify the effectiveness of the presented work.
|
|---|
