Comparing performance of multiple classifiers for regression and classification machine learning problems using structured datasets

Abstract

Abstract Machine learning aims at improving performance through repeated trials using different classifiers. Given the large number of classification algorithms and difference in their output on account of differing underlying models, it’s important to formulate guidelines on which classification algorithm to use under specific conditions. This study examines the performance of different classifiers on datasets in the public domain and presents a comparison of the results obtained. Based on results found for multiple datasets XG Boost AS is seen to produce superior or comparable results when measured in terms of linear regression, standard deviation and mean absolute error found in the test data making it a candidate for consideration for regression as well as classification machine learning problems especially for structured data.