Multiobjective Optimization of Classifiers by Means of 3-D Convex Hull Based Evolutionary Algorithm

TL;DR

This paper extends a convex hull-based multiobjective evolutionary algorithm to handle higher-dimensional classification problems, including multi-class and complexity objectives, demonstrating its effectiveness on benchmarks and real-world tasks.

Contribution

It introduces a novel extension of the convex hull-based genetic programming algorithm for higher-dimensional multiobjective classification problems.

Findings

Effective in capturing relevant convex hull regions

Demonstrates robustness on benchmark datasets

Useful in real-world email classification and feature selection

Abstract

Finding a good classifier is a multiobjective optimization problem with different error rates and the costs to be minimized. The receiver operating characteristic is widely used in the machine learning community to analyze the performance of parametric classifiers or sets of Pareto optimal classifiers. In order to directly compare two sets of classifiers the area (or volume) under the convex hull can be used as a scalar indicator for the performance of a set of classifiers in receiver operating characteristic space. Recently, the convex hull based multiobjective genetic programming algorithm was proposed and successfully applied to maximize the convex hull area for binary classification problems. The contribution of this paper is to extend this algorithm for dealing with higher dimensional problem formulations. In particular, we discuss problems where parsimony (or classifier complexity) is stated as a third objective and multi-class classification with three different true classification rates to be maximized. The design of the algorithm proposed in this paper is inspired by indicator-based evolutionary algorithms, where first a performance indicator for a solution set is established and then a selection operator is designed that complies with the performance indicator. In this case, the performance indicator will be the volume under the convex hull. The algorithm is tested and analyzed in a proof of concept study on different benchmarks that are designed for measuring its capability to capture relevant parts of a convex hull. Further benchmark and application studies on email classification and feature selection round up the analysis and assess robustness and usefulness of the new algorithm in real world settings.