An Explainable Approach to Land-Use Classification Using CNN-LSTM and LIME

The evolution of landscapes has opened the opportunity to utilize previously unused land for various purposes, such as agriculture and urban development. This approach promotes sustainable land management and addresses economic and societal needs. By repurposing land effectively, growth and sustainability in communities can be fostered. This study introduces a hybrid CNN-LSTM model for classifying satellite images in the EuroSAT dataset, encompassing land use categories. The Convolutional Neural Network (CNN)efficiently extracts spatial features, while the Long Short-Term Memory (LSTM) captures temporal patterns, achieving an impressive accuracy of 98.30%. Preprocessing includes grayscale conversion, contrast enhancement, and spatial-temporal augmentations to optimize data quality. Explainable AI, implemented via Local Interpretable Model-agnostic Explanations (LIME), provides insights into influential features guiding predictions and enhances interpretability.