Clustering of Large Document Sets with Restricted Random Walks on Usage Histories

An update algorithm for restricted random walk clustering for dynamic data sets

“…single linkage clustering, our tests in the context of the OPAC usage data set indicate that it still produces very large clusters with up to 200,000 objects, which makes it unsuited for applications like recommender services (Franke and Thede 2005). This effect is mainly due to bridge elements in the data.…”

“…The problematic aspect of the step measure is that for instance the fifth step from a 20-step walk receives the same importance as the fifth step from a 5-step walk, although the latter carries more importance due to its relative position. Franke and Thede (2005) propose the relative measures of the level l + = step number walk length+1 and l − = step number−1 walk length which are asymptotically equivalent since their difference converges to zero.…”

“…The reformulation of the second-order Markov chain shown here is due to Franke and Thede (2005). It is based on an irreducible, infinite, first order Markov chain defined on the edges of the graph.…”

“…In the purchase data set, this corresponds to changes in cross-selling numbers as the purchase data base grows. Both cases occur within the OPAC data set described in Franke and Thede (2005), therefore only algorithms that can cope both with changing data set sizes and dynamic similarities can be used in the scenario. Similar requirements hold for services in social network sites.…”

Using Restricted Random Walks for Library Recommendations

“…Technically, a recommender service can be implemented in different ways. We will present an innovative approach based on a fast clustering algorithm for large object sets [2] and making use of product cross-occurrences in purchase histories: In our case, the purchase histories are those of users of the Online Public Access Catalogue (OPAC) of the university's library at Karlsruhe, and a purchase is the viewing of a document's detail page in the WWW interface of the OPAC. A cross-occurrence between two documents is given when their detail pages have been viewed together in one user session.…”