Large Experiment and Evaluation Tool for WEKA Classifiers.
Dustin D. Baumgartner, Gürsel Serpen · 2009
Abstract- This paper presents a new Windows®-based software utility for WEKA, a data mining software workbench, to simplify large-scale experiment and evaluation with many algorithms and datasets in the classification context. The proposed tool, LEET (Large Experiment and Evaluation Tool) makes it possible to accomplish a variety of tasks that are presently rather difficult or impractical through the standard WEKA interfaces. This includes allowing comparison of classifiers across multiple experiments, tracking execution time, calculating diversity measures, and summarizing the characteristics of many datasets. We have tested and validated LEET as part of a study with 50+ machine learning classification/ensemble algorithms, 46 datasets, and calculation of a variety of performance measures. With WEKA providing the algorithm implementations, LEET facilitates the execution and evaluation of large-scale experiments with greater ease than any existing interface.