OPTIMIZING PREDICTIVE ACCURACY: A COMPARATIVE ANALYSIS OF SUPERVISED MACHINE LEARNING ALGORITHMS FOR REAL-WORLD CLASSIFICATION TASKS

Abstract

Supervised machine learning methods are widely used for binary classification tasks; however, their effectiveness depends on data characteristics and model assumptions. This study presents a comparative evaluation of commonly used supervised learning algorithms, including Logistic Regression, Support Vector Machine, K-Nearest Neighbors, Decision Tree, Random Forest, and Naïve Bayes. The analysis is conducted on a balanced dataset consisting of 400 samples with seven predictive features and a binary target variable. A unified experimental framework is applied to ensure fair comparison, and model performance is assessed using accuracy, precision, recall, F1-score, and Receiver Operating Characteristic Area Under the Curve (ROC–AUC). Visual evaluation techniques, including ROC and precision–recall curves, support interpretability. Experimental results indicate that ensemble and kernel-based models outperform linear and probabilistic approaches. In particular, the Random Forest classifier achieves the highest performance, demonstrating strong accuracy, balanced classification, and superior discriminative capability. The findings provide practical guidance for supervised model selection in structured binary classification problems.