Research Background and motivation

With the growing complexity of the society, consumer preference has become more unpredictable because of more sources of information collection and more products choices. To enhance the capability of competition on the market, to know customer more has always the most important mission to companies. In marketing research, the explanatory value of traditional criteria is in steady decline. This is because the individuals who together make up the market every day provide examples of very similar purchase behavior patterns on the part of people differing considerably in social-economic and demographic terms and vice versa, with growing personalization of consumer habits being observed (Gonzalez and Laurentino, 2002). Therefore, the deep and wide-ranging changes which present-day society is undergoing must also be taken into account.

This study will separate into two parts according to the different channels that consumer can shop through them, that is, virtual channel and physical channel. The former is the channel that consumer should shop by going to the “brick and mortar”

store, the latter is about internet online shopping. There are many lifestyle researches have analyzed consumer behavior of different product targets in the physical channel, for instance, food, red wine or trip choices. However, the lifestyle approach those researches used mostly are product-specification, which means the usage of the research is limited, i.e., the results of the analysis will be hard to apply to other product categories later. And also, their model of doing the marketing research has its

own drawbacks. For instance, questionnaire is often too long to make sure respondents provide the right answers or the research method could miss or neglect huge amount of information form the data which is collected.

To solve the problems, this research aims to improve the traditional research model, through literature review and expert advice. A general lifestyle structure is going to be generated, after the appropriate methodologies are applied, the patterns of consumer behavior will be discovered, and that is, the decent linkages between consumer’s lifestyle and their product choice can be found. After obtaining the result of analysis, we can deduce the marketing strategies and suggestions to the companies.

Instead of physical channel, virtual channel such as internet has become an important way of shopping nowadays. Internet provides new opportunities for sellers and buyers. With the increasing number of people shop thought this channel, the market have enjoyed a rapid growth. Business-to-consumer electronic commerce is growing in every category of goods, for instance, financial services, online-travel services, computer hardware and software, book and music all accompanied with good sales performance in these years (Kim et al., 2001). However, due to the characteristics of the e-commerce industry, companies with limited resources face extremely high competition, to discover effective means of marketing online will always be a critical issue to internet stores. To achieve the maximum efficiency of marketing, the key to success will rely on the market segmentation, that is, the important fundamental marketing analytic basis first proposed by Wendell (1956).

Since companies need to satisfy the diverging preferences of customers and react to the complex online business environment quickly, the new marketing strategies such as one-to-one marketing have been stressed by researches and practical affairs. One to one marketing (also known as database marketing) introduces a fundamental new

basis for competition in the marketplace by enabling organizations to differentiate based on customers rather than products (Peppers and Rogers, 1993). The main purpose of one to one marketing is not market share but finding a group of valuable customers toward specific products. One solution to realize these strategies is personalized recommendation that helps customers find the product according to their interests by producing a list of products for each given customer (Cho et al., 2002).

That is, an effective way to increase customer satisfaction and consequently customer loyalty.

Nowadays, a variety of recommendation techniques has been developed and recommendation systems are commonly applied by B to C companies, for instance, Amazon.com and CDNow.com, they use an intelligent engine to mine the customers＇ ratings records and then create predictive user models for product recommendation. Typically there are two kinds of recommender systems:

content-based and collaborative systems. The former provide recommendations to a customer by automatically matching customers＇ interests with product contents; the latter, has been known to be the most successful, provide recommendations by utilizing overlap of preference ratings to combine the opinions of “like-minded＂

customers, that is, identify customers whose interests are similar to those of a given customer have liked.

Until now, there are researches about one to one marketing, that is, using data mining techniques to segment the market and forming the recommendation systems use demographics variable as the only basis of market segmentation, that is, lifestyle and values are not involved in the research. And also, many lifestyle researches about online shopping; however, most of their target is to find a wired-lifestyle, that is, using life-related variables to analyze if people shop online or not. Concerning the description above, to add lifestyle and values as variables to segment the market could

enrich the recommendation systems. Therefore, to increase the accuracy of matching the customer and products online, the general lifestyle structure is worth to be included. Since the variables used to segment the market of recommend system are not only demographics, the understanding of customers can be improved and the efficiency of the interaction with them will be increased hopefully.

In this study, questionnaire design requires special data mining tools which and deal with different data scales, considering several methodologies, rough set theory (RST) and classification and regression tree (CART) are used in this study to analyze the content and features of data.

Classification and regression tree, developed by (Breiman et al., 1984), is a flexible and robust analytical method, which can deal with nonlinear relationships, high-order interactions, missing values, and this method is simple to understand and give easily interpretable results. There are several advantages of CART: (1) the flexibility to handle a broad range of response types, including numeric, categorical, ratings, and survival data; (2) invariance to monotonic transformations of the explanatory variables; (3) ease and robustness of construction; (4) ease of interpretation; and (5) the ability to handle missing values in both response and explanatory variables. Thus, trees complement or represent an alternative to many traditional statistical techniques, including multiple regression, analysis of variance, logistic regression, log-linear models, linear discriminant analysis, and survival models. Rough set theory, which was developed by Pawlak (1982), is a rule-based decision-making technique that can handle crisp datasets and fuzzy datasets without need for a pre-assumption membership function. It can also deal with uncertain, vague, and imperceptible data.

The remainder of this study is organized as follows. Section 2 describes the literature review to be the fundamental of the study. Section 3 describes the methodology of classification and regression tree and rough set theory. In section 4,

two real cases such as the purchasing behavior of digital camera and books online are presented to show the new segmentation process by CART and RST. Finally, in section 5, we present the conclusions and suggestions.