بررسی دانش محصول خاص کاربران برای شخصی سازی در تجارت الکترونیک

While many electronic commerce (EC) companies are adopting one-to-one marketing approaches using various personalization technologies to make their products and services unique for the purpose of attracting and retaining customers and improving their completion edges in the EC ecosystem, which, nevertheless, has low entrance barriers for new players to join and further intensify the competition, none or few of them consider a fundamental issue—the user's product-specific knowledge. Our research proposed to add this new domain of the customer's knowledge on appropriate target products into the personalization process as a part of the overall EC strategy for businesses. In this paper, we present our initial design for assessing the user's product-specific knowledge using the proposed innovative method for detecting it directly in a non-intrusive way without asking users to answer or fill out any types of questionnaires. Our method is based on customer's on-line navigation behaviors by analyzing their navigation patterns through pre-trained artificial neural networks. An empirical study designed for a case of EC store selling digital cameras was conducted in our research to prove the concept, and a good preliminary result was derived from the study. For the purpose of comparing the performances between the conventional approach of using questionnaire and the proposed innovative approach of navigation pattern mining, a questionnaire based approach for evaluating the user's product-specific knowledge was designed and incorporated into our knowledge level assessment system (KLAS). Our study result shows that although the pure questionnaire-based KLAS is intrusive and may not be accepted by some users, for those users willing to complete the questionnaire, the proposed navigation pattern approach can be combined with the questionnaire-based approach to create a hybrid KLAS which has a significantly improved accuracy rate in detecting the customer's product knowledge level.

In recent years, people's life and living styles have been deeply influenced by Internet, which enables electronic commerce (EC) for companies and their business partners to conduct business and perform electronic transactions (Lin and Lu, 2000 and Liu and Arnett, 2000). In addition to the purchase of products and services over the Internet, EC also encompasses all electronically conducted business activities, operations, and transaction processing within and cross companies. Through EC, companies can alleviate constraints (upon time, space, and cost) to enhance the way they connect to and interact with their EC counterparties by serving customers and collaborating with business partners electronically and intelligently. To catch the revolutionary opportunity and benefit of EC, an explosive number of companies are competing in the EC ecosystem, which, nevertheless, has low entrance barriers for new players to join and further intensify the competition. Thus, for the purpose of attracting and retaining customers and then improving their completion edges, some EC companies take advantage of differentiation and personalization technologies to make their products and services unique and to tailor their products and services for specific user preferences. For example, through personalization, businesses can research on customer's behaviors for developing appropriate marketing strategies, and then delivering suitable products and services to the targeted customers accordingly. Wind and Rangaswamy (2001) found that the opportunity and capability to offer consumers a flexible and personalized relationship is probably one of the most important advantages among all possible benefits offered by EC to businesses. It is shown that personalization can ultimately enhance customer's satisfaction level and loyalty, and the increase in each customer's visiting frequency can further create more transaction opportunities and benefit the Internet businesses (Lee, Liu, & Lu, 2002). From the consumers' point of view, the Internet has become a major channel to the worldwide sources of information. While the Internet traffic has been increasing rapidly since 1997, at the range between 70 and 150% annually (Odlyzko, 2003), it is estimated that the amount of information available from Internet doubled every 18 months, and the number of home pages is even increasing in a faster rate (Yang, Yen, & Chen, 2000). This fact is causing users a serious problem of information overload when they try to retrieve information from the dynamically and continuously growing web resources. Therefore, the need from web users in identifying and using more intelligent systems or tools for conducting information gathering and information filtering from the huge size of web related sources is on the rise (Li & Zhong, 2004). In regard to products and services, different people have different and/or various needs, interests, and preferences; nevertheless, the taste and inclination of a person on products and services may also change or evolve with time. Thus, the ‘one-to-one marketing’ strategy was proposed to provide personalized service in the EC environment (Allen et al., 1998 and Weng and Liu, 2004). ‘If we have two million customers, then we should have two million shops on our website’, said by Jeff Bezos, CEO of Amazon, may serve as an example to show the importance and the value of the personalization strategy in EC environment. Personalization technology can give users a better, in terms of efficiency and effectiveness, EC experience since they do not have to browse through all the irrelevant noise. In general, there are two major approaches to provide personalized information: content-based and collaborative filtering (Aggarwal et al., 1999 and Yu, 1999). In the content-based approach, it matches the content of candidate items against the user profile, which is constructed by analyzing the content of items that the user has favored in the past or user's personal information and preferences. Some recommendation systems, which are used by EC companies to suggest products and provide information to customers, operate based on this approach, such as NewsWeeder (Lang, 1995) and Infofinder (Krulwich & Burkey, 1996). In the collaborative filtering approach, it identifies other users that have showed similar preference to the given users and provides what they would like. Several recommendation systems are developed based on this approach, such as Tapestry (Goldberg, Nichols, Oki, & Terry, 1992), GroupLens (Konstan, Miller, Maltz, Herlocker, Gordon, & Riedl, 1997), Ringo (Shardanand & Maes, 1995), PHOAKS (Terveen, Hill, Armento, McDonald, & Creter, 1997), and Siteseer (Rucker & Polenco, 1997). While the content-based personalization suffers limitations in dealing with non-text multimedia resources (such as movies, music, etc.) and in making classified recommendations other than the localized domain specified by a user's profile/preference, the collaborative filtering approach is unable to provide new items to a user and unsuited to a user with changing or evolving preferences. Various renovated or hybrid approaches were proposed to cope with the shortcomings of content-based personalization and collaborative filtering, and to increase the accuracy of recommendation systems, by integrating content-based approach and collaborative filtering approach (Changchien et al., 2004 and Weng and Liu, 2004), applying data mining techniques (such as association rules mining) to collaborative filtering (Kim et al., 2002, Lee et al., 2001, Wang and Shao, 2004 and Wang and Thao, 2003), and combining collaborative filtering based on item and collaborative filtering based on user approaches (Li, Lu, & Xuefeng, 2005). Although all the above-mentioned recommendation systems share the same spirit of assisting in the user's search of items of interest, none of them address a fundamental issue—the user's product-specific knowledge. Our research proposed to add this new domain of customer's knowledge on all potential products into the personalization process as a part of the overall EC strategy for businesses. In this paper we present our initial design for assessing the user's product-specific knowledge. Since a user's product-specific knowledge level on various products varies, we proposed an innovative method for detecting it directly in a non-intrusive way without asking the user to answer or fill out any types of questionnaires. Our method is based on the customer's on-line navigation behaviors by analyzing their navigation patterns through pre-trained artificial neural networks. An empirical study designed for a case of EC store selling digital cameras was conducted in our research to prove the concept, and a good preliminary result was derived from the study. This automatic and non-intrusive approach for evaluating the customer's product-specific knowledge was incorporated into our personalized promotion decision support system (Changchien et al., 2004), which used data mining techniques in accordance with marketing strategies to help the business prepare the highly potential and suitable promotion products for each individual customer. The subsequent sections of this article are organized as follows. Section 2 describes product knowledge, web usage mining, and back propagation networks. Section 3 proposes a method together with its two variations for evaluating the customer's product-specific knowledge level. Section 4 shows our experiment results, and Section 5 concludes this paper after the discussions.

It is quite a challenge that a business faces more competitors in Internet than in traditional market, and the customer's loyalty in the Internet is low compared with traditional market so that it is a difficult problem for a business to attract and retain customers in EC. Traditional mass marketing is no longer effective for EC in the Internet, and thus more precise on-line one-to-one marketing for better suiting each customer becomes more and more important for competing on the Internet, along with the use of highly advanced data analysis techniques and the development of new marketing strategies for EC. Hence, an on-line personalized promotion decision support system (PPDSS) was developed in our previous research (Changchien,et al., 2004), to assist a business in intelligently developing the marketing strategy and the on-line one-to-one promotion products based on the experiences analyzed and retrieved from the historical transactions. In this paper, an innovative non-intrusive method of assessing the customer's product knowledge level, which can be used to enhance the effectiveness of personalization strategy in an EC environment, was proposed and implemented through our knowledge level assessment system (KLAS). Since KLAS was designed to ameliorate the constrains of the log based and the agent based web usage mining techniques by using a server-side approach to track and record users' browsing behaviors on the website, it could be more effectively and proactively incorporated into various applications in gaining competing advantage for EC companies. For the purpose of evaluating the performance of our proposed system, we also enhanced KLAS to accept questionnaire-based input so that not only could the results from various approaches be compared quantitatively, but the navigation pattern based KLAS (see Fig. 3) and the questionnaire-based KLAS (see Fig. 5) could be combined to derive a hybrid KLAS (see Fig. 6). It was shown from our experiments that the proposed navigation pattern based KLAS achieved an accuracy rate of 78% in detecting whether a user is an expert or not, the benchmark derived from the questionnaire-based KLAS offered a better accuracy rate of 84%, and the hybrid KLAS delivered the best accuracy rate of 93%. It is good to see that the proposed navigation pattern based KLAS is a promising way of dynamically, proactively, and non-intrusively detecting users' product-specific knowledge levels, given that its accuracy rate derived from our experiments is good and only slightly lower than the accuracy rate of the questionnaire-based KLAS, which, however, is intrusive and may not be accepted by at least some users. However, for those users willing to complete the questionnaire, the proposed navigation pattern approach can be integrated with the questionnaire-based approach to create a hybrid KLAS which can achieve a significantly improved accuracy rate (93% based on our experiments). In addition to learning and constructing customer profiles and preferences (i.e. to know about the users), EC businesses may use KLAS to further enhance their personalization strategy by understanding the customer's knowledge level about each target product (i.e. to know about what the users know about the product). This domain knowledge of the customer's product-specific knowledge level can be applied to various areas (such as customer segmentation, customer relationship management, call center, web personalization, recommendation system, etc.) in conducting EC business. For example, in a one-to-one marketing website, the information content, information organization, and information presentation about a promotion item can be tailored according to various knowledge levels of customers on the target item. Knowing a user's product-specific knowledge level has the potential to confer considerable competing advantage in the development of ‘dynamic personalization’ strategy for EC companies. The capability and methodology of assessing user's product-specific knowledge quickly and effectively can facilitate the design of an EC website for dynamic personalization, which in turn can be used to facilitate the customer orientation design. Customer orientation always emphasizes on customer's interest, and it stresses the derivation of customer profiles and the construction of customer knowledge base for enterprises as an intangible asset that is difficulty to be imitated by competitors (Deshpande, Farley, & Webster, 1993). Through the customer orientation strategy to leverage customer knowledge, enterprises can avoid competing on pricing and flexibly provide differential prices based on the customer's demand curve, and eventually get higher average prices (Roberts, 2000). Our proposed KLAS based dynamic personalization is a feasible and attractive tool for today's enterprises to enhance the effectiveness of customer orientation and create strategic advantage by raising the entrance barriers. To prove the concept, not only was KLAS integrated into our PPDSS to show how KLAS can achieve the goal of dynamic personalization and customer orientation design (see Fig. 2), but an empirical study was conducted using a KLAS application website for selling digital cameras. Although we have proposed an innovative, non-intrusive, and promising method for assessing users' product-specific knowledge, there are some constraints on its general applicability. First, each implementation of this method would only work on one category of EC products. For an EC website selling various categories of products, the complexity and the degree of difficulty in design and implementation of the KLAS system would significantly increase. Second, KLAS relies on its BPN model for detecting whether a user is an expert, and the BPN model needs to be trained before KLAS can be put in use. However, the BPN training process requires pre-defined and mutually exclusive expert patterns and non-expert patterns, as well as a training dataset collected in advance. The accuracy of KLAS is highly dependent on the quality and the quantity of the training dataset. Third, this method may only work for products with some specific characteristics and attributes. Since digital camera was selected as the sample product in our study, it is highly possible that KLAS can be applied to other 3C products, but it is still uncertain whether KLAS can be used by websites selling non-3C products. Some pretests or further empirical studies may need to be conducted to classify the candidate products before incorporating KLAS into an EC website, and it is clear that a consistent measurement for product categorization needs to be established for using the proposed KLAS.