Institutional Repository
Technical University of Crete
Technical University of Crete
EN
|
EL
| URI: | http://purl.tuc.gr/dl/dias/3205D4B3-6DF2-42EA-8137-727E4C9E7BD3 | ||
| Year | 2018 | ||
| Type of Item | Diploma Work | ||
| License |
|
||
| Bibliographic Citation | Θεόδωρος Φλώκος, "Ανάπτυξη πολυκριτήριας μεθοδολογίας υπολογισμού βαρών σε ensemble τεχνικές μηχανικής μάθησης", Διπλωματική Εργασία, Σχολή Μηχανικών Παραγωγής και Διοίκησης, Πολυτεχνείο Κρήτης, Χανιά, Ελλάς, 2018 https://doi.org/10.26233/heallink.tuc.78696 | ||
| Appears in Collections |
The need for expert knowledge fusion and it’s use for decision support, is a troubling subject for the modern decision sciences. One of the problems the experts face during the process of decision making is the classification of objects (ex: products, services, conditions etc.) in multiple classes. On the other hand, the «explosion» of data in the modern era , makes the classification of objects a difficult task for experts. To tackle this set of problems, numerous computational models in the field of machine learning have been developed, that simulate the way experts make decisions using empirical data. In the context of classification those models are called «classifiers» . For the post optimization of the decision making process a class of methods have been developed for the fusion of the classifiers which are called «ensemble» methods. One of the common problems in the process of creating an ensemble is defining the distribution of weights of the classifiers. Most of ensemble methods set the classifiers weights based on the classifiers performance or they set equal weights for all the classifiers. Though the classification performance measures show some advantages and disadvantages depending the hypothesis each of them is based on. So the selection of single classification performance measure to define the classifiers weights is not an effective technique. We are going to model such a problem as multicriteria problem, taking into account a set of classification performance measures. The goal of this dissertation is to develop an «ensemble» method that takes into account all of the things mentioned above in the definition of classifiers weights, with the help of multicriteria decision analysis.
Copyright © DIAS 2013
