یک مدل تصادفی از رفتار الکترونیکی مشتری

Web usage mining techniques are increasingly used today to understand e-customers’ within-site behavior. We propose a data mining model that considers e-customers’ activities as a discrete-time semi-Markov process and explains their behavior. An algorithm is proposed to compute transition probability matrix and holding time mass functions from the site navigation data. Finally, the model is used to explain customer behavior in an example site. A software agent, implemented in the site, collects and stores navigation data in the required form and thus helps to avoid data preprocessing. The model results helped to improve the site design and judge its performance.

Websites having emerged as the first contact point with the customers for many organizations, the study of e-customer’s behavior has become an essential field of research. In a traditional business environment, when a seller meets a buyer in person he can understand his behavior, like intention for purchase and choice of product, etc., from his expressions and verbal communication. The sales person accumulates this knowledge while dealing face to face with the customers. Subsequently, he uses this knowledge to create facilities to increase customer satisfaction. But, in an online environment, no such knowledge can be directly obtained. Online feedback forms can help to get customer preferences. Customer preferences can be analyzed to understand e-customer behavior. But there is no way to ensure that a visitor fills up this form. An alternative is to conduct visitor surveys. However, unless properly designed, these surveys may involve samples from specific parts of the visitor population and may not depict a true picture of customer behavior due to the global nature of the ecommerce environment. In this scenario, the website visitors’ page access pattern can be analyzed, using the Web usage mining techniques, to understand the e-customers’ within-site behavior. Analysis of e-customer behavior can help in improving the content and design of the website, customizing the website, building stronger customer relationship, heightening communication with the customer, and in enhancing customer service. The rest of the paper is organized as follows. First, we give a comprehensive survey of literature in the area of e-customer behavior. Thereafter we propose a model of e-customer behavior treating it as a semi-Markov process. Next, we present an algorithm that transforms the navigational data to a form that can be readily used by the model. Finally, we use the model to analyze the navigational data of an example site and use the results to find a few design defects in the site and predict the performance of the site

In this paper we forward semi-Markov process model as a Web usage mining tool to understand e-customer behavior. We have developed an algorithm to find out the transition probability matrix and holding time mass functions from the site visitors’ navigational data. The model is implemented in an example site. A software agent is implemented in the site to collect and store navigation data in the required form. This mode of data collection helps to bypass the tedious data preprocessing steps. The model results have helped in improving the design of the example site. They have also helped in judging the extent to which the purpose of the site has been achieved. A few difficulties may be encountered while implementing the proposed model. The performance of the model to correctly predict the visitors’ behavior depends on the accuracy of the navigational data collected from the Web server. The page caching by the clients browser and the proxy server is a major barrier for successful data collection. It also depends on the proper categorization of the pages contained in the site into various states. The categorization may change with addition or deletion of pages in a site. Further, it varies from site to site and depends on the Web master’s personal judgment and bias. The estimated holding times include the network latency and server-side processing times, and hence they do not give a clear picture of the times spent by a visitor in the state. The work presented in this paper can be extended in many ways. If the visitors can be classified into various groups then the analysis of the generated data can give better insights into the behavior of that group. The transition matrix captures a compact history of users’ navigational behavior and can be used to customize the website by pre-fetching information relevant for a new visitor. Another extension of the present work is to dynamically generate hyperlink structure of a website. The algorithm developed by us for dynamic hyperlink generation has been published elsewhere [51]. Such a dynamic link customization can help to guide a visitor during his tour within the site.