Scientific Understanding, Experimental Analysis and a Survey on Evolution of Classification Rule Mining Based on Ant Colony Optimization

Given the explosive rate of data deposition on the web; classification has become a complex and dynamic phenomenon. As classification complexity is continuing to grow, so is the need in direct proportion to designing and developing data mining algorithms & techniques. Classification is the most commonly applied data mining technique, a process of finding a set of models or functions that describes and distinguishes data classes, for the purpose of using it – so classification is a specialist with specialized skills, which is moving toward universality. A classification problem is considered as a supervised learning problem. The aim of the classification task is to discover a kind of relationship between the attributes (input) and class (output), so that the discovered knowledge can be used to predict the class of a new unknown object. Classification of the records or data is done based on the classification rules. Ant colony optimization is a method that derives its inspiration from real ants that forage for food by selecting the shortest path from multiple possible paths available to reach food. Thus merging the concept of Ant Colony Optimization (ACO) with data mining brings in a new approach to designing classification rule that will be helpful in extraction of information for a specialized dataset. In this paper a survey is done on Ant-miner algorithm for classification Rule extraction. The Ant miner algorithm extract classification rule from data using if-then-else pattern; similar to other traditional algorithm available for classification task or purposes. Extraction of classification Rule from data is an important task of data mining. We present, detailed description about the algorithm available for classification rule mining using Ant colony optimization. Variations to the ant colony based an Ant-miner algorithm is discussed along with the comparison of the algorithms with critical parameters like predictive accuracy, No. of Rules Discovered, No. of terms per No. of rules Discovered, using different data sets. Hence the paper will help to study various ant miner algorithms and comparison carried out will help the data miner to select and use algorithm according to need based on the specialized properties associated with the algorithm.