|
|
Decision Trees for Classification
Čechmánek, Jakub ; Štanclová, Jana (advisor) ; Štefka, David (referee)
There is a lot of approaches for data classification problems resolving. The most significant data classification methods are neural networks, Bayes nets, clustering, linear models, associative rules, etc. This thesis deals with decision trees which deserves attention of experts as well. Step by step are discussed C4.5, CART and SDT trees, a variant of classical decision tree inductive learning using fuzzy sets theory. Substantial part of work is devoted to pruning algorithms as well. Particular methods are examined and compared over freely available data sets of feature vectors with respect to stopping criteria, splitting criteria of a node and size of constructed trees. A summary of our own results is included.
|
|
|
Decision Trees for Classification
Čechmánek, Jakub ; Štefka, David (referee) ; Štanclová, Jana (advisor)
There is a lot of approaches for data classification problems resolving. The most significant data classification methods are neural networks, Bayes nets, clustering, linear models, associative rules, etc. This thesis deals with decision trees which deserves attention of experts as well. Step by step are discussed C4.5, CART and SDT trees, a variant of classical decision tree inductive learning using fuzzy sets theory. Substantial part of work is devoted to pruning algorithms as well. Particular methods are examined and compared over freely available data sets of feature vectors with respect to stopping criteria, splitting criteria of a node and size of constructed trees. A summary of our own results is included.
|
guest ::
