OPTIMISASI HIPERPARAMETER ALGORITMA MACHINE LEARNING MENGGUNAKAN GRID SEARCH UNTUK KLASIFIKASI KELANGSUNGAN HIDUP PENUMPANG KAPAL

Abstract

Passenger survival prediction provides a relevant classification setting for evaluating the ability of machine learning algorithms to identify risk patterns from passenger characteristics. This study aims to evaluate the effects of hyperparameter optimisation on classification performance and to compare the effectiveness of Grid Search, Random Search, and Bayesian Optimisation in identifying optimal model configurations. The Titanic dataset was employed as a case study, with the analytical workflow comprising data preprocessing, missing-value treatment, categorical variable encoding, normalisation, and class balancing using the Synthetic Minority Over-sampling Technique (SMOTE). Three machine learning algorithms—K-Nearest Neighbours (KNN), Logistic Regression, and Support Vector Machine (SVM)—were evaluated using accuracy, precision, recall, F1-score, and the Area Under the Receiver Operating Characteristic Curve (AUC–ROC). The results demonstrate that hyperparameter optimisation improves model performance and consistency, with KNN and SVM exhibiting more pronounced improvements, whereas Logistic Regression shows comparatively limited changes in performance. Comparison across optimisation approaches indicates that Grid Search can produce competitive hyperparameter configurations through systematic exploration of a predefined parameter space. These findings demonstrate that algorithm selection and hyperparameter optimisation strategy are important components in developing accurate and reliable classification models. Methodologically, the proposed approach provides a potential foundation for developing data-driven predictive models to support risk analysis and decision-support applications in the maritime transport sector.