Supervised Learning Algorithms and Evaluation Metrics in Machine Learning

Authors

  • Bharath Ganji
  • Riyaz Hussain Shaik
  • Satya Prakash G

Abstract

In the era of emerging technologies, advancements in machine learning techniques have created a great impact in our lives. These advancements made us to develop new way of approaches to solve problems in different areas such as cancer diagnosis, predictive forecasting, speech recognition etc. Various machine learning techniques like Supervised learning, Unsupervised learning, Reinforcement learning have been extensively used to transform a computer into an intelligent machine to solve complex and challenging problems in real world. The advancements in our technologies in terms of computational power and acquiring large data made machine learning models to grow complex day-by-day. So building a good model is not sufficient since evaluating a model is equally important. So a good metric is needed to estimate the prediction error. Although plethora of metrics are available in the community of machine learning, confusion arises very often in choosing the right metric. However there is no common theory in choosing a metric to evaluate a model. This paper describes popular supervised learning algorithms along with the mathematics behind them. We analyze and evaluate the important evaluation metrics in classification and regression algorithms. After assessing the discussed evaluation metrics, a conclusion is made on choosing the right evaluation metric for the given problem.

Downloads

Published

2020-01-18

Issue

Section

Articles