Improvement of Classification Accuracy in Machine Learning Algorithm by Hyper-Parameter Optimization

Abstract

The manual optimization of hyperparameters is a straightforward and well-known approach, but it is not scalable, particularly when there are several settings and options. In nearly every area of daily life, machine learning offers more logical guidance than humans can. It has already been noted in the literature that correct Hyper-Parameter optimization has a significant impact on a machine learning algorithm’s performance. Manual search is one method for performing Hyper-Parameter optimization, however it takes a lot of time. Some of the common techniques used for hyperparameter optimization include grid search, random search, and optimization procedure. The main model training and structural hyper-parameters are introduced in the first part, along with their significance and approaches for defining the value range. The research then concentrates on the main optimization techniques and their applicability, examining their effectiveness and accuracy, particularly for the random forest ensemble algorithm. In this study, we present a novel approach for enhancing the Random Forest algorithm’s hyperparameters using the Parkinson’s Disease Data Set. Accuracy, precision, recall and F1 score were taken into account while comparing the performances of each of these strategies.