In this paper we discuss an interesting and useful property of clickstream data. Often a visit includes repeated views of the same page. We show that in three real datasets, sampled from the websites of technology and consulting groups and a news broadcaster, page repetitions occur for the majority as a very specific structure, namely in the form of nested palindromes. This can be explained by the widespread use of features which are available in any web browser: the "refresh" and "back" buttons. Among the types of patterns which can be mined from sequence data, many either stumble if symbol repetitions are involved, or else fail to capture interesting aspects related to symbol repetitions. In an attempt to remedy this, we characterize the palindromic structures, and discuss possible ways of making use of them. One way is to pre-process the sequence data by explicitly inserting these structures, in order to obtain a richer output from conventional mining algorithms. Another application we discuss is to use the information directly, in order to analyze certain aspects of the website under study. We also provide the simple linear-time algorithm which we developed to identify and extract the structures from our data.

Mining web traffic data has been addressed in literature mostly using sequential pattern mining techniques. Recently, a more powerful pattern called partial order was introduced, with the hope of providing a more compact result set. A further approach towards this goal, valid for both sequential patterns and partial orders, consists in building a statistical significance test for frequent patterns. Our method is based on probabilistic generative models and provides a direct way to rank the extracted patterns. It leaves open the number of patterns of interest, which depends on the application, but provides an alternative criterion to frequency of occurrence: statistical significance. In this paper, we focus on the construction of an algorithm which calculates the probability of partial orders under a first-order Markov reference model, and we show how to use those probabilities to assess the statistical significance of a set of mined partial orders.

In this paper we address the following question: how to estimate a Markov Decision Process modeling the dynamics of customer relationships. Once the model is estimated, we discuss how to efficiently allocate marketing resources and instruments in order to maximize the long-term value generated by customers in a given future time horizon using dynamic programming. Our methodology allows us both to predict and to optimize the future value generated by customers. We show our approach using a case study involving a major European airline.

This paper addresses the problem of image classification using local information which is aggregated to provide global representation of different image classes. Local information is adaptively extracted from an image database using Independent Component Analysis (ICA) which provides a set of localized, oriented, and band-pass filters selective to independent features of the images. Local representation using ICA techniques has been previously investigated by several researchers. However, very little work has been done on further use of these representations to provide more complex and global description of images. In this paper, we present an algorithm which uses the energy of a minimal set of ICA filters to provide class-specific signatures which are shown to be strongly discriminant. Computer simulations are carried on two image databases, one consisting of five classes--referred to as categories--(buildings, rooms, mountains, forests and beaches) and one consisting of a set of 30 objects from multiple views for viewpoint invariant object recognition. The classification performance of the algorithm using both Independent and Principal Component Analyses are reported and discussed.