We present a new methodology for exploring and analyzing navigation patterns on a web site. The patterns that can be analyzed consist of sequences of URL categories traversed by users. In our approach, we rst partition site users into clusters such that users with similar navigation paths through the site are placed into the same cluster. Then, for each cluster, we display these paths for users within that cluster. The clustering approach weemployis model-based (as opposed to distance-based) and partitions users according to the order in which they request web pages. In particular, we cluster users by learning a mixture of rst-order Markov models using the Expectation-Maximization algorithm. The runtime of our algorithm scales linearly with the number of clusters and with the size of the data � and our implementation easily handles hundreds of thousands of user sessions in memory. In the paper, we describe the details of our method and a visualization tool based on it called WebCANVAS. We illustrate the use of our approach on user-tra c data from msnbc.com. Keywords: Model-based clustering, sequence clustering, data visualization, Internet, web 1

We explore opportunities for targeted pricing for a retailer that only tracks weekly storelevel aggregate sales and marketing-mix information. We show that it is possible, using these data, to recover essential features of the underlying distribution of consumer willingness to pay. Knowledge of this distribution may enable the retailer to generate additional profits from targeting by using choice information at the checkout counter. In estimating demand we incorporate a supply-side model of the distribution channel that captures important features of competitive price-setting behavior of firms. This latter aspect helps us control for the potential endogeneity generated by unmeasured product characteristics in aggregate data. The channel controls for competitive aspects both between manufacturers and between manufacturers and a retailer. Despite this competition, we find that targeted pricing need not generate the prisoner’s dilemma in our data. This contrasts with the findings of theoretical models due to the flexibility of the empirical model of demand. The demand system we estimate captures richer forms of product differentiation, both vertical and horizontal, as well as a more flexible distribution of consumer heterogeneity.

We present a mixture model based approach for learning individualized behavior models for the Web users. We investigate the use of maximum entropy and Markov mixture models for generating probabilistic behavior models. We first build a global behavior model for the entire population and then personalize this global model for the existing users by assigning each user individual component weights for the mixture model. We then use these individual weights to group the users into behavior model clusters. We show that the clusters generated in this manner are interpretable and able to represent dominant behavior patterns. We conduct offline experiments on around two months worth of data from CiteSeer, an online digital library for computer science research papers currently storing more than 470,000 documents. We show that both maximum entropy and Markov based personal user behavior models are strong predictive models. We also show that maximum entropy based mixture model outperforms Markov mixture models in recognizing complex user behavior patterns. 1. Introduction and Related

There are now available a number of new subscription services that comprise a dual pricing system of a monthly access fee (rental) and a per-minute usage charge. Examples include cellular phones, the Internet, and pay TV. The usage and retention of such services depend on the absolute and relative prices of this dual system. For instance, a moderate access fee but a low-usage charge might initially appeal to customers, but later a low-usage customer might find the monthly fee unjustified and thereby relinquish the service. Providers of such services, therefore, usually offer several pricing packages to cater to differing customer needs. The purpose of this study is to derive a revenue-maximizing strategy for subscription services, that is, the combination of access and usage price that maximizes revenue over a specified

Feature subset selection is important not only for the insight gained from determining relevant modeling variables but also for the improved understandability, scalability, and possibly, accuracy of the resulting models. Feature selection has traditionally been studied in supervised learning situations, with some estimate of accuracy used to evaluate candidate subsets. However, we often cannot apply supervised learning for lack of a training signal. For these cases, we propose a new feature selection approach based on clustering. A number of heuristic criteria can be used to estimate the quality of clusters built from a given feature subset. Rather than combining such criteria, we use ELSA, an evolutionary local selection algorithm that maintains a diverse population of solutions that approximate the Pareto front in a multi-dimensional objective space. Each evolved solution represents a feature subset and a number of clusters; two representative clustering algorithms, K-means and EM, are applied to form the given number of clusters based on the selected features. Experimental results on both real and synthetic data show that the method can consistently find approximate Pareto-optimal solutions through which we can identify the significant features and an appropriate number of clusters. This results in models with better and clearer semantic relevance. 1.

This paper evaluates complete versus partial classification for the problem of identifying customers at risk. We define customers at risk as customers reporting overall satisfaction, but these customers also possess characteristics that are strongly associated with dissatisfied customers. This definition enables two viable methodological approaches for identifying such customers, i.e. complete and partial classification. Complete classification entails the induction of a classification model to discriminate between overall dissatisfied and overall satisfied instances, where customers at risk are defined as overall satisfied customers who are classified as overall dissatisfied. Partial classification entails the induction of the most prevalent characteristics of overall dissatisfied customers in order to discover overall satisfied customers who match these characteristics. In our empirical work, we evaluate complete and partial classification techniques and compare their performance on both quantitative and qualitative criteria. The intent of the paper is not on proving the superiority of partial classification, but rather to provide an alternative and valuable approach that offers new and different insights. In fact, taking predictive accuracy as the performance criterion, results for this study show the superiority of the complete classification approach. On the other hand, partial classification offers additional insights that complete classification techniques do not offer, i.e. it offers a rule-based description of criteria that lead to dissatisfaction for locally dense regions in the multidimensional instance space.

File sharing is one of the most controversial applications in the Internet. Millions of users enjoy downloads of billions of media files such as songs or movies. But where do these files come from? While the economic rationale to download files is obvious, the motives of individuals to actually share and, therefore, internalize the costs (i.e. the risk of being sued for copyright infringement) are less obvious. Consequently, empirical studies have shown a large proportion of users demanding files but not offering any and, therefore, free riding on their peers. Nevertheless, sharing can be rational. This paper offers a theoretical base to explain sharing behavior and proves that the users ’ utility considerations depend on the network’s life cycle. At the beginning of a network’s life cycle the incentives to share files are high if an individual understands the economics of network externalities. Nonetheless, the user’s utility to share files decreases over time, especially if the network grows and becomes anonymous. The strategies of individual users to share or free ride are explained using game theoretic approaches. Depending on the life cycle, the expectations and utilities of the users differ, leading to various games and optimal strategies. We empirically test our hypotheses using mixture regression models and explain the rationale of sharers in different stages of the life cycle. In order to prevent the (theoretically) inevitable break-down of the file sharing network, the authors finally present strategies for file sharing networks to enhance the user’s willingness to share.

This study is dedicated to proposing a two-stage method, which first uses Self-Organizing Feature Maps (SOM) neural network to determine the number of clusters and cluster centroids, then uses honey bee mating optimization algorithm based on K-means algorithm to find the final solution. The results of simulated data via a Monte Carlo study show that the proposed method outperforms two other methods, SOM followed by K-means (Kuo, Ho & Hu, 2002a) and SOM followed by GAK (Kuo, An, Wang & Chung, 2006), based on both within-cluster variations (SSW) and the number of misclassification. In order to further demonstrate the proposed approach’s capability, a real-world problem of an internet bookstore market segmentation based on customer loyalty is employed. The RFM model is used for comparison of customers ' loyalty. Then the proposed method is used to cluster the customers. The results also indicate that the proposed method is better than the other two methods.

informs ® doi 10.1287/mksc.1050.0161 © 2006 INFORMS The recent marketing literature reflects a growing interest in structural models, stemming from (1) the desire to test a variety of behavioral theories with market data, and (2) recent developments that facilitate estimation of and inference for these models. Whether one should always go through the effort of developing such tightly parameterized models with the associated computational burden of estimating them and whether it pays off to make strict behavioral assumptions in terms of better decisions remain open questions. To shed some light on these issues, we provide examples of structural approaches to consumer choice and demand as well as examples where the goal is to study the nature of competition in the marketplace. From that review comes our discussion of issues in the development and application of structural models, including their estimation, testing, and validation, their applicability in the practice of marketing, and their usefulness for normative as well as descriptive purposes. Key words: structural models; heterogeneity; competition; endogeneity; dynamic demand models. History: This paper was received September 14, 2004, and was with the authors 3 months for 2 revisions; processed by Kannan Srinivasan. 1.

: In this paper the "Self-Organizing (Feature) Map" (SOM) methodology as originally proposed by Kohonen (1982) is employed in the context of Competitive Market Structure (CMS) and segmentation analysis using household-specific brands preferences derived from diary panel data as input patterns for SOM training. The adaptive SOM algorithm results in a representation of competitive structures among rival brands at the segment-level, i.e. for submarkets with (internally) more homogeneous brand choice features. This property of SOM-based CMS/segmentation analysis allows the two interdependent tasks to be performed simultaneously, as it is frequently claimed in the marketing literature. 1 Introduction Competitive Market Structure (CMS) analysis refers to the task of deriving a configuration of brands in a product class on the basis of their competitive relationships (cf. DeSarbo et al.(1993)). Among the numerous approaches introduced into marketing literature, it is widely accepted to opera...