Experiencing KEEL software: Application to fatigue data segment classification

Abstract

Most researchers have doubts over what type of classification model is most suitable for representing their data when applying a data mining approach. However, this quandary can be solved by performing a comparison between the available models. Generating multiple classification models simultaneously is cost-effective when using free data mining tools such as KEEL (Knowledge Extraction based on Evolutionary Learning). KEEL is an interactive data mining software that provides a high level of functionality for users. This paper presents an experience of using this user-friendly software to solve a fatigue data segment classification problem. The problem concerns classifying fatigue data segments whether they consist of low or high damage. Experiments over the dataset using three different types of classification algorithms (i.e. CN2, C4.5 and Naïve-Bayes), two discretization methods, and two filter-type feature selection methods show that there is no optimal classification algorithm in terms of indexing performance for our specific problem. Interestingly, however, this study does provide minor evidence that classifier performance could be increased by pairing discretization and feature selection methods appropriately. © 2014 AIP Publishing LLC.