A comparison of Finite State Classifier and Mahalanobis-Taguchi System for multivariate pattern recognition in skin cancer detection

Abstract

This project presents two methods for image classification for the detection of malignant melanoma: the Mahalanobis-Taguchi System and Finite State Classifiers. The Mahalanobis-Taguchi System is a diagnosis and predictive method for analyzing patterns in multivariate cases, while Finite State Classifiers are a state based machine learning technique. The goal of this study is to compare the ability of the Mahalanobis-Taguchi System and a Finite State Classifier to discriminate using small data sets. We examine the discriminant ability as a function of data set size using publicly available skin lesion image data. While analysis of the data shows a high degree of correlation, the Mahalanobis-Taguchi System performed poorly when trying to discriminate between Malignant Melanoma and benign lesions. Alternately, the Finite State Classifiers developed using evolutionary computation obtained over 85% correct classification of the malignant and benign lesions using the image data sets.