"Nen\u00ED k dispozici"@cs . "Macek, Jan" . "Nen\u00ED k dispozici"@cs . . "2004-04-13+02:00"^^ . "Vienna" . "Vienna" . "Austrian Society for Cybernetics Studies" . "3-85206-169-5" . . "Cybernetics and Systems 2004" . . "Gaussian Complexities Based Decision Tree Pruning" . "RIV/68407700:21230/04:03096739" . . . "Gaussian Complexities Based Decision Tree Pruning"@en . "Lhotsk\u00E1, Lenka" . "4"^^ . "Z(MSM 210000012)" . . . "2"^^ . . . "21230" . "2"^^ . "565193" . "Gaussian Complexities Based Decision Tree Pruning"@en . . . . "We investigate a method for pruning of decision trees based on the complexity measure of a tree and its error rate. It uses the Gaussian complexity averages of a decision tree to estimate the error rate of classification. This complexity measure is data-dependent, so we expect it to allow the pruning to capture the dependencies in the data better than error based pruning methods or general complexity based pruning methods. We perform experiments that compare unpruned decision tree, decision tree pruned with our method and decision tree pruned with Reduced Error Pruning method. Our results show that proposed method outperforms both mentioned methods" . "Decision Tree Pruning; Gaussian Comlexity Measure"@en . "[05166F46FB28]" . "Gaussian Complexities Based Decision Tree Pruning" . . . "We investigate a method for pruning of decision trees based on the complexity measure of a tree and its error rate. It uses the Gaussian complexity averages of a decision tree to estimate the error rate of classification. This complexity measure is data-dependent, so we expect it to allow the pruning to capture the dependencies in the data better than error based pruning methods or general complexity based pruning methods. We perform experiments that compare unpruned decision tree, decision tree pruned with our method and decision tree pruned with Reduced Error Pruning method. Our results show that proposed method outperforms both mentioned methods"@en . "719 ; 722" . . "Nen\u00ED k dispozici"@cs . . "RIV/68407700:21230/04:03096739!RIV/2005/MSM/212305/N" .