Decision trees construction and rules extraction from trained neural networks

The problems of sharing use of neural network technologies with logical deduction and decision-making methods in data mining are considered. The analysis of existing algorithms and methods of decision trees creation is carried out. Algorithms for decision tree construction based on neural networks and rules extraction from trained neural networks are presented. One of the directions to use decision trees for data mining is to extract rules from trained neural networks. For this purpose, a series of algorithms could be applied, in particular NeuroRule class, performing thinning of the network and identifying the most significant features.