Mining Numerical Association Rules Using Genetic Algorithm

Abstract

Since the introduction, association rule mining technique became the most famous and widely used data mining technique because of its simplified nature of solution that it generates. Because of its robustness of deriving associations among various attributes of dataset it is used in various application areas for decision making, detection and prediction etc. Although the technique seems to be very easy in starting, especially when dealing with categorical data but becomes quite complex when it’s time to deal with numeric data because of its diversity. This work is done to deal with problem of association rule mining from numeric data in an efficient way and aimed to generate more interesting rules from the dataset. For accomplishing this task we have used a well-known machine algorithm i.e. genetic algorithm as the base of the solution to this problem. Genetic algorithm is selected for this task because of its nature of self-improving and ability to handle large solution set. Here we have proposed two algorithm based on genetic algorithm with slight differences. These algorithms are also implemented and tested on various datasets and a comparison between proposed and existing work has also been illustrated.