Optimalisasi ANN-MLP Dengan Grid Search-CV Untuk Klasifikasi Tutupan Lahan Perkotaan Menggunakan Sentinel-2

Abstract

Efficient urban land cover classification is crucial for urban planning and management. This study aims to optimize the Artificial Neural Network Multi-Layer Perceptron (ANN-MLP) model using GridSearchCV for urban land cover classification in Padang City. Sentinel-2 satellite imagery served as the primary data source. The research methodology encompassed image data Preprocessing, Feature Extraction, dataset partitioning into Training and Testing sets, ANN-MLP model Training , and hyperparameter tuning using GridSearchCV. The results demonstrated that the ANN-MLP optimized with GridSearchCV achieved an overall accuracy of 98.33% and a Kappa value of 98%. This research underscores the effectiveness of GridSearchCV in determining the optimal hyperparameter configuration and the importance of visual map evaluation to identify limitations and opportunities for further improvement in Sentinel-2-based land cover analysis.