Tuning the hyperparameters of multilayer perceptron and support vector machines classifiers in binary classification problems

In machine learning, it is generally aimed to create a model that works well and makes accurate predictions. Optimization of parameter values provides a good solution for this purpose. The aim of this thesis study is to perform parameter optimization to increase the success of Support Vector Machines and Multilayer Perceptron machine learning algorithms in binary classification problems. For this, the contributions of Random Search, Bayesian Search, and Optuna optimization methods to these classifiers were handled. In this context, experimental studies were carried out on three datasets. For both Support Vector Machines and Multilayer Perceptron, Random Search and Bayesian Search methods mostly outperformed the Optuna method. Moreover, experimental studies showed us that optimization methods can have a positive effect on these classifiers depending on the problem under consideration. KEYWORDS: Support vector machines, multi-layer perceptron, binary classification, optimization, hyperparameter tuning. June 2022, 55 Page,