Relational Markov models (RMMs) are a generalization of Markov models where states can be of different types, with each type described by a different set of variables. The domain of each variable can be hierarchically structured, and shrinkage is carried out over the cross product of these hierarchies. RMMs make effective learning possible in domains with very large and heterogeneous state spaces, given only sparse data. We apply them to modeling the behavior of web site users, improving prediction in our PROTEUS architecture for personalizing web sites. We present experiments on an e-commerce and an academic web site showing that RMMs are substantially more accurate than alternative methods, and make good predictions even when applied to previously-unvisited parts of the site.

Predictive Web prefetching refers to the mechanism of deducing the forthcoming page accesses of a client based on its past accesses. In this paper, we present a new context for the interpretation of Web prefetching algorithms as Markov predictors. We identify the factors that affect the performance of Web prefetching algorithms. We propose a new algorithm called WM o , which is based on data mining and is proven to be a generalization of existing ones. It was designed to address their specific limitations and its characteristics include all the above factors. It compares favorably with previously proposed algorithms. Further, the algorithm efficiently addresses the increased number of candidates. We present a detailed performance evaluation of WM o with synthetic and real data. The experimental results show that WM o can provide significant improvements over previously proposed Web prefetching algorithms.

... this article, we examine this issue by classifying and discussing a wide ranging set of Web metrics. We present the origins, measurement functions, formulations and comparisons of well-known Web metrics for quantifying Web graph properties, Web page significance, Web page similarity, search and retrieval, usage characterization and information theoretic properties. We also discuss how these metrics can be applied for improving Web information access and use.

Visitors who browse the web from wireless PDAs, cell phones, and pagers are frequently stymied by web interfaces optimized for desktop PCs. Simply replacing graphics with text and reformatting tables does not solve the problem, because deep link structures can still require minutes to traverse. In this paper we develop an algorithm, MINPATH, that automatically improves wireless web navigation by suggesting useful shortcut links in real time. MINPATH finds shortcuts by using a learned model of web visitor behavior to estimate the savings of shortcut links, and suggests only the few best links. We explore a variety of predictive models, including Na ve Bayes mixture models and mixtures of Markov models, and report empirical evidence that MINPATH finds useful shortcuts that save substantial navigational effort. 1

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

This paper examines the accuracy of predicting a user's next action based on analysis of the content of the pages requested recently by the user. Predictions are made using the similarity of a model of the user's interest to the text in and around the hypertext anchors of recently requested Web pages. This approach can make predictions of actions that have never been taken by the user and potentially make predictions that reflect current user interests. We evaluate this technique using data from a full-content log of Web activity and find that textual similarity-based predictions outperform simpler approaches.

The primary goal of Web usage mining is the discovery of patterns in the navigational behavior of Web users. Standard approaches, such as clustering of user sessions and discovering association rules or frequent navigational paths, do not generally provide the ability to automatically characterize or quantify the unobservable factors that lead to common navigational patterns. It is, therefore, necessary to develop techniques that can automatically identify the users' underlying navigational objectives and to discover hidden semantic relationships among users as well as between users and Web objects. Probabilistic Latent Semantic Analysis (PLSA) is particularly useful in this context, since it can uncover latent semantic associations among users and pages based on the co-occurrence patterns of these pages in user sessions. In this paper, we develop a unified framework for the discovery and analysis of Web navigational patterns based on PLSA. We show the flexibility of this framework in characterizing various relationships among users and Web objects. Since these relationships are measured in terms of probabilities, we are able to use probabilistic inference to perform a variety of analysis tasks such as user segmentation, page classification, as well as predictive tasks such as collaborative recommendations. We demonstrate the e#ectiveness our approach through experiments performed on several real-world data sets.

The analysis of the hyperlink structure of the web has led to significant improvements in web information retrieval. This survey describes two successful link analysis algorithms and the state-of-the art of the field.

Predicting the next request of a user as she visits Web pages has gained importance as Web-based activity increases. Markov models and their variations, or models based on sequence mining have been found well suited for this problem. However, higher order Markov models are extremely complicated due to their large number of states whereas lower order Markov models do not capture the entire behavior of a user in a session. The models that are based on sequential pattern mining only consider the frequent sequences in the data set, making it di#cult to predict the next request following a page that is not in the sequential pattern. Furthermore, it is hard to find models for mining two di#erent kinds of information of a user session. We propose a new model that considers both the order information of pages in a session and the time spent on them. We cluster user sessions based on their pair-wise similarity and represent the resulting clusters by a click-stream tree. The new user session is then assigned to a cluster based on a similarity measure. The click-stream tree of that cluster is used to generate the recommendation set. The model can be used as part of a cache prefetching system as well as a recommendation model.

We propose mixtures of hidden Markov models for modelling clickstreams of web surfers. Hence, the page categorization is learned from the data without the need for a (possibly cumbersome) manual categorization. We provide an EM algorithm for training a mixture of HMMs and show that additional static user data can be incorporated easily to possibly enhance the labelling of users. Furthermore, we use prior knowledge to enhance generalization and avoid numerical problems. We use parameter tying to decrease the danger of over tting and to reduce computational overhead. We put a at prior on the parameters to deal with the problem that certain transitions between page categories occur very seldom or not at all, in order to ensure that a nonzero transition probability between these categories nonetheless remains. In applications to arti cial data and real-world web logs we demonstrate the usefulness of our approach. We train a mixture of HMMs on arti cial navigation patterns, and show that the correct model is being learned. Moreover, we show that the use of static 'satellite data' may enhance the labeling of shorter navigation patterns. When applying a mixture of HMMs to realworld web logs from a large Dutch commercial web site, we demonstrate that sensible page categorizations are being learned.