AN ARTIFICIAL NEURAL NETWORK CLASSIFICATION APPROACH FOR IMPROVING ACCURACY OF CUSTOMER IDENTIFICATION IN E-COMMERCE

With the advancesin Web-based oriented technologies, experts are able to capture user activities on the Web. Users' Web browsing behavior is used for user identification. Identifying users during their activities is extremely important in electronic commerce (e-Commerce) as it has the potential to prevent illegal transactions or activities particularly for users who enter the system through the use of unknown methods. In addition, customer behavioral pattern identification provides a wide spectrum of applications such as personalized Web pages, product recommendations and present advertisements. In this research, a framework for users' behavioral profiling formation is presented and customer behavioral patternsare used for customer identification in the e-Commerce environment. Based on activity control, policies such as user restriction or blockingcan be applied. The neural network classification and the measure of similarity among behavioral patterns are two approaches applied in this research. The results of multi-layer perceptron with a back propagation learning algorithm indicate that there is less error and up to 15.12% more accuracy on average. The results imply that the accuracy of the neural network approach in customer pattern behavior recognition increases when the number of customers grows. In contrast, the accuracy of the similarity of pattern method decreases.