Optimization of Binary Classification Based on Receiver Operating Characteristic Area Under the Curve for Supervised Machine Learning

Receiver Operating Characteristic - Area Under the Curve (ROC_AUC) is widely used to see the performance of an algorithm. In this research, the use of ROC_AUC is to increase the optimality of the accuracy of the results in binary classification. The method used is to select features that have an AUC score above 0.5. The results of these features are used for the binary classification process that is applied to lowdimensional datasets (Wine and Loan datasets) and highdimensional datasets (Parkinson dataset). The results showed an increase in the accuracy of the K-Nearest Neighbor (KNN) and Decision Tree (DT) algorithms for loan datasets and Stochastic Gradient Descent (SGD) for the Wine and Loan dataset. As for the Parkinson's dataset, the accuracy of the MLP algorithm increases. On the Loan dataset, the KNN algorithm increased its accuracy from 59 % to 76%, SGD increased its accuracy from 57 % to 70 %, and DT increased its accuracy from 63 % to 71 %. In the Wine dataset, the SGD algorithm has increased its accuracy from 36 % to 93 %. In the Parkinson's dataset, the Multi-layer Perceptron (MLP) algorithm has increased its accuracy from 65 % to 72 %.