By Carmela Cappelli, Francesco Mola (auth.), Prof. Dr. Hans-Hermann Bock, Prof. Marcello Chiodi, Prof. Antonino Mineo (eds.)
This quantity features a number of papers provided through the biennial assembly of the type and information research workforce (CLADAG) of the Societa Italiana di Statistica which was once orga nized via the Istituto di Statistica of the Universita degli Studi di Palermo and held within the Palazzo Steri in Palermo on July 5-6, 2001. For this convention, and after checking the submitted four web page abstracts, fifty four papers have been admitted for presentation. They lined a wide range of themes from multivariate information research, with detailed emphasis on category and clustering, computa tional information, time sequence research, and functions in a variety of classical or fresh domain names. A two-fold cautious reviewing method resulted in the choice of twenty-two papers that are awarded during this vol ume. they impart both a brand new notion or technique, current a brand new set of rules, or challenge a fascinating software. we've clustered those papers into 5 teams as follows: 1. class tools with purposes 2. Time sequence research and similar equipment three. machine in depth options and Algorithms four. type and knowledge research in Economics five. Multivariate research in technologies. In every one part the papers are prepared in alphabetical order. The editors - of them the organizers of the CLADAG confer ence - wish to exhibit their gratitude to the authors whose enthusiastic participation made the assembly attainable and intensely successful.
Read or Download Advances in Multivariate Data Analysis: Proceedings of the Meeting of the Classification and Data Analysis Group (CLADAG) of the Italian Statistical Society, University of Palermo, July 5–6, 2001 PDF
Similar analysis books
This recommendations handbook is aimed at teachers to be used as a significant other quantity to the ebook, a contemporary thought of Integration (AMS Graduate stories in arithmetic sequence, quantity 32).
- Asymptotics for Orthogonal Polynomials
- Zahlentheoretische Analysis II
- Multiple Criteria Decision Analysis: State of the Art Surveys
- Analysis 1
Additional resources for Advances in Multivariate Data Analysis: Proceedings of the Meeting of the Classification and Data Analysis Group (CLADAG) of the Italian Statistical Society, University of Palermo, July 5–6, 2001
In this paper we analyze features and limitations of two proximity measures between classification trees: one is a distance that measures the amount of rearrangement needed to change one of the trees so that it has structure identical to the other (Shannon and Banks (1999)) , while the other is a similarity measure that compares the partitions associated to the trees taking into account their relative predictive power (Miglio (1996)) . We suggest a normalizing factor for the distance defined by Shannon and Banks; furthermore , we propose a new dissimilarity measure that considers both the aspects explored separately in the previous ones.
When two classification trees have to be compared, both aspects (the structure and the predictive power) should be simultaneously considered. In fact, trees having the same distance with respect to their structures can show a very different predictive power. On the other hand, trees with the same predictive power can have very different structures. For this reason, we propose a new dissimilarity measure, which considers the structure and the predictive power at the same time. Its definition is the following: (5) 32 Miglio and Soffritti where Sih and Sjk are similarity coefficients whose values synthesize the similarities Shk between the H leaves of Ti and the K leaves of Tj , computed as follows: Shk = mhkChk JmhOmOk , h = 1, ...
00028 160 - 27814 190 -5 - 166982 Situation 2 Fig. 3. Filtered images two sit uat ions, ranging from mod erat e to great difficulty of classificat ion. 100 replications were obtained with identical st atisti cal properti es. The perform ance of t he classifier was measured through the error rat e of classification, which was est imated by th e Mont e Carlo method. The est imated error rat e of classification corresponds to th e average of th e error rat es found for t hese replications. 27% for each considered sit- 24 De Carvalho et al.
Advances in Multivariate Data Analysis: Proceedings of the Meeting of the Classification and Data Analysis Group (CLADAG) of the Italian Statistical Society, University of Palermo, July 5–6, 2001 by Carmela Cappelli, Francesco Mola (auth.), Prof. Dr. Hans-Hermann Bock, Prof. Marcello Chiodi, Prof. Antonino Mineo (eds.)