Nell'ambito del Dottorato di Ricerca in Statistica Metodologica presso il Dip. di Statistica, Probabilitą e Statistiche Applicate, Univ. di Roma "La Sapienza"
Il giorno 16 Ottobre 2001 alle ore 15.00 presso la Sala 34
il prof. Joe Whittaker (University of Lancaster)
terrą un seminario dal titolo: "A graphical models view point of some topics of multivariate statistics"
Nell'ambito del Dottorato di Ricerca in Statistica Metodologica presso il Dip. di Statistica, Probabilitą e Statistiche Applicate, Univ. di Roma "La Sapienza"
Il giorno 20 Ottobre 2001 alle ore 10.30 presso la Sala 34
il prof. Jan Larsen (University of Denmark)
terrą un seminario dal titolo: "Mining the Web"
Abstract:
Automated analysis
of the world wide web is a new challenging area
relevant in many applications, e.g., retrieval, navigation and organization of information, automated information assistants, and e-commerce. The talk
will discuss hierarchical methods for unsupervised and supervised web mining
which provide
multilevel description of data. In particular I focus on agglomerative
probabilistic
clustering from Gaussian density mixtures with new probabilistic
similarities
measures. An unique advantage of using the probabilistic
clustering scheme
is automatic
detection of the final hierarchy level for new data not used for training.
In order to provide
a meaningful description of the clusters we further suggest two
interpretation
techniques: listing of prototypical data examples from the cluster, and listing
of
typical features
associated with the cluster.
The techniques will
be demonstrated on various web mining applications: classification of web
pages
and segmentation of
emails. |