Relating statistical methods to machine learning predictive models

Abstract

The paper reviewed the probabilistic feature of binomial distribution in the operation of machine learning (ML) classifications. It also examined a normal distribution and the concepts for approximating the binomial distribution to a normal distribution in estimating generalization error and its role in machine learning model selection. Again, it studied the confident interval and hypothesis testing and their estimations in the evaluation and comparison of the Performance metrics (Accuracy) of the learning algorithms. The paper highlighted their statistical significance to the ML models and classifiers as well as the differences in their utilization in statistics and machine learning.