ارزیابی عملکرد کمک به مشتری فروشگاه های آنلاین همراه با مدل های عامل مبتنی بر رفتار مشتری و استراتژی تکامل

With competitive pressure growing in online markets, many Internet stores provide various customer aid functions such as personalized pages to help customers shop more effectively and efficiently. Evaluating such customer aid functions is usually costly because it requires full or partly-working systems and many human testers. In order to address this problem, this research presents a novel approach to evaluating customer aid functions with agent-based models of customer behavior and evolution strategies. Agent-based modeling is used to imitate users’ rational behavior at Internet stores with regard to browsing and collecting product information. It is assumed that users evolve their browsing skill and strategy over time, to maximize the efficiency and effectiveness of their shopping, and hence, evolution strategy, an optimization method, is combined with the agent-based model to find the rational behavior of each user. The rational behavior is then used to simulate the virtual shopping of users and to evaluate the performances of target customer aid functions. Several experiments were performed to illustrate the use of the approach, where the personalized recommendation page of a virtual online DVD rental store is evaluated in comparison with more general functions such as listing most popular products or sorting categories. The results show that a personalized page might not always be the best customer aid function for all users compared to the simpler ones.

With e-commerce maturing across industries and countries, being able to sell products or services online no longer gives much competitive advantage to businesses. Any company is now able to quickly build an e-commerce infrastructure and launch services with readily available technology solutions and the use of development experts at low cost. One of the efforts that many companies have made in order to survive in this competitive environment is to develop various customer aid functions that can provide a more satisfying shopping experience to customers when compared to rival Internet stores. There are many such customer aid functions that range from simply assorting products in various ways to providing intelligent personalized services. For example, many stores such as Amazon.com recommend products in several different ways, according to individual customers’ activities or transaction history [33]. Many stores also have personalized pages that provide a list of recommended products, information, or advertisements that are prepared according to the estimated preference of each customer, based on their activities or transaction records. There are also simpler and non-personalized recommendations provided in many stores that show the lists of best selling products, top-rated products, most-clicked products, etc. Simple customer aid functions are relatively easy to build, but sophisticated methods such as personalization often require advanced techniques based on statistics or artificial intelligence. The main goal of this research is to develop a novel method of analyzing and evaluating the usefulness of such customer aid functions by developing an agent-based model of customer behavior and applying evolution strategy, an optimization technique, to the model. Therefore, this research first develops an agent-based model of individual customer’s behavior with parameters that can represent the different behavioral characteristics of individuals. Then, evolution strategy is used to identify the optimal or rational behavior of each customer by optimizing the parameters of the agent-based model for each shopper. The results of the optimization are used for simulating the behavior of each customer and the simulation results are analyzed to evaluate the usefulness of target customer aid functions. Compared with many other methods or guidelines for evaluating user interfaces or web personalization methods, there are several advantages to this approach for evaluating customer aid functions. First, the approach attempts an experimental analysis without requiring the implementation of customer aid functions or expensive empirical studies involving many human participants [26], [52] and [56]. Similarly, the approach is also different from many software testing methods that mostly focus on finding errors and defects in software, where again the implementation of such software is needed, along with the participation of human testers .[9] and [50]. Second, while it is usually difficult to quantify many general measures of design principles or web-site evaluation frameworks for the automated evaluation of customer aid functions [11], [20] and [54], this approach enables the simulated analysis of customer aid functions using direct measures of effectiveness and efficiency. In order to illustrate this approach, a series of experiments were performed using a virtual online DVD rental store, composed of publicly available datasets. In the experiments, the usefulness of some customer aid functions was evaluated in comparison with others. The paper is organized as follows. Section 2 gives a brief overview of the related literature. Section 3 presents the agent-based model of customer behavior and shows the application of evolution strategy to the model. Section 4 shows the results of the experiments. Section 5 discusses the results followed by the presentation of the conclusion in Section 6.

The contributions of this research can be summarized as follows. First, to the author’s knowledge, this is the first piece of research that shows how agent-based modeling and evolution strategy can be used together for evaluating customer aid functions at Internet stores. Second, the suggested method can be used to evaluate customer aid functions under various assumptions at much lower cost compared with previous attempts based on expensive tests or empirical experiments involving human participants with fully or partly-working systems. Third, compared with most other studies on recommendation systems that only measure how accurate the estimation of customer preferences is for products, the method showed how customer performances can be measured directly when using personalized recommendations. Fourth, the application of the approach to the example of the online DVD store showed a very interesting result which indicates that sophisticated functions such as personalized pages might not always be helpful for customers with regard to improving their purchasing process. Although this research showed the model of customer behavior applied to a single product type in the form of a virtual Internet store, the method can also be modified and extended for analyzing businesses in other industrial sectors. However, care should be taken in generalizing the findings of the analysis to other areas because different types of customers or products may exhibit disparate shopping behavior, leading to very different conclusions. Although there are some limitations to the study and further research issues remaining, the author believes that the method, if applied properly, can facilitate the development and evaluation of many creative new customer aid functions for a wide variety of Internet stores.