RFE-Based Feature Selection to Improve Classification Accuracy for Morphometric Analysis of Craniodental Characters of House Rats

Abstract

In conventional morphometrics, researchers often collect and analyze data using large numbers of morphometric features to study the shape variation among biological organisms. Feature selection is a fundamental tool in machine learning which is used to remove irrelevant and redundant features. Recursive feature elimination (RFE) is a popular feature selection technique that reduces data dimensionality and helps in selecting the subset of attributes based on predictor importance ranking. In this study, we perform RFE on the craniodental measurements of the Rattus rattus data to select the best feature subset for both males and females. We also performed a comparative study based on three machine learning algorithms such as Naïve Bayes, Random Forest, and Artificial Neural Network by using all features and the RFE-selected features to classify the R. rattus sample based on the age groups. Artificial Neural Network has shown to provide the best accuracy among these three predictive classification models.