Mining for Appraisal of Knowledge extracted from Realistic Data

Abstract

The meaning of the traditional machine intelligence or artificial intelligence (knowledgebase systems and expert systems that depend on the knowledge represented in high level language) has been broadened and improved by the field of softcomputing. Due to the complexities in data, extracting knowledge out of it, intelligently, is one of the greatest challenges in the field of AI. Several computational techniques or algorithms have been used to deal with such complex and nonlinear systems. The Nature and biology inspired computational algorithms i.e. neural networks (human-brain inspired), evolutionary computation (human-evolution inspired), and fuzzy logic (linguistic representation) take ideas from natural and biological processes inorder to develop several artificial systems that can efficiently adapt to the continually changing environment. However such techniques work properly and the desired result is met under certain conditions. As far as Fuzzy Logic is concerned the major issue lies in building up of membership functions and making decision on its appropriate parameters (MFs, their distribution and composition of fuzzy rules) and that in Neural Network, the architecture is an important issue. So taking such issues into consideration, the combined system of the two is used and the optimization technique (evolutionary computation) is applied over it to optimize various parameters like number of membership functions for each input. Such softcomputing techniques prove to be faster than the individual ones. So, our main reason behind using computational techniques is to act in similar way as biological process solves our real life computational issue. The basic principle of working lies on adaptability based on computational system wherein the advantages of the individual techniques are focused on and appropriate parameters are selected optimally that can result in less error and better accuracy.