Classification of two-level data with hyperrectangles parallel to the coordinate axes
Subject Areas : عمومىzahra moslehi 1 , palhang palhang 2
1 -
2 -
Keywords:
Abstract :
One of the learning methods in machine learning and pattern recognition is supervised learning. In supervised learning and in two-category problems, the available educational data labels include positive and negative categories. The goal of the supervised learning algorithm is to calculate a hypothesis that can separate positive and negative data with the least amount of error. In this article, among all supervised learning algorithms, we focus on the performance of decision trees. The geometric view of the decision tree brings us closer to the concept of separability in computational geometry. Among all the available resolution algorithms related to the decision tree, we raise the problem of calculating the rectangle with the maximum difference of two colors and implement the algorithm in one, two, three and m dimensions, where m represents the number of data features. The implementation result shows that this algorithm is competitive with the well-known C4.5 algorithm.