Optimalisasi ANN-MLP dengan GridSearch-CV untuk Klasifikasi Tutupan Lahan Perkotaan Menggunakan Sentinel-2

Sihaloho, Mayhendra Daud
Yulfa, Arie

Geadidaktika
Vol 5, No 2 (2025): Geadidaktika Agustus 2025

Accurate land cover classification is essential for sustainable urban planning and management. This study optimizes the Artificial Neural Network Multi-Layer Perceptron (ANN-MLP) model using GridSearchCV and Sentinel-2 imagery to classify urban land cover in Padang City. Based on 500 samples across five land cover classes and validated with high-resolution imagery, the optimized model achieved 97% accuracy and a Kappa value of 96.25%. These results highlight the effectiveness of hyperparameter optimization in improving classification performance while offering practical contributions for local governments, including mapping urban growth, identifying land-use changes, guiding development according to environmental capacity, and strengthening data-driven spatial planning policies. The proposed approach can also be replicated in other regions with similar characteristics.