A Random Forest Classification Algorithm Based on Dichotomy Rule Fusion
Yueyue Xiao, Wei Huang, Jinsong Wang
Abstract
The classical random forest algorithm has associated features and bias problems, which leads to a reduction in classification accuracy, in this paper we propose a random forest classification algorithm based on dichotomy rule fusion. The dichotomy rule fusion method is based on the idea of information gain and recursive feature elimination to select a better feature sequence, which improves the classification accuracy. Experimental results on international standard data sets show that the algorithm has better performance in classification than some commonly used algorithms.
Topics & Concepts
Random forestFeature (linguistics)Computer scienceStatistical classificationFusionArtificial intelligencePattern recognition (psychology)AlgorithmMajority ruleRule-based systemInformation fusionSequence (biology)Classification ruleData miningMachine learningPhilosophyBiologyLinguisticsGeneticsFace and Expression RecognitionRough Sets and Fuzzy LogicText and Document Classification Technologies