Komparasi Teknik Hyperparameter Optimization pada SVM untuk Permasalahan Klasifikasi dengan Menggunakan Grid Search dan Random Search

Abstract

Classification is one of the important tasks in the field of Machine Learning. Classification can be viewed as an Optimization Problem (Optimization Problem) with the aim of finding the best model that can represent the relationship/pattern between data with labels. Support Vector Machine (SVM) Is an algorithm in Machine Learning used to solve problems such as Classification or Regression. The performance of the SVM algorithm is strongly influenced by parameters, for example error prediction in non-linear SVM results in parameters C and gamma. In this study, an analysis of the technique was carried out to obtain good parameter values using Grid search and Random Search on seven datasets. Evaluation is done by calculating the value of accuracy, memory usage and validity test time of the best model by the two techniques.