Interpretable Bi-LSTM Based Model for Understanding Urban Growth Patterns Using Explainable AI

Urban growth has emerged as one of the most daunting planning issues in rapidly expanding metropolitan regions, spurred by an intricate interdependence of socio-economic, environmental, and spatial considerations. In this investigation, it is suggested that an interpretable deep learning model that combines Bidirectional Long Short-Term Memory (Bi-LSTM) networks with Explainable AI (XAI) methods be used to model and analyze urban growth patterns. Taking advantage of a high-dimensional Sustainable Urban Planning and Landscape dataset, the model is preprocessed with state-of-the-art preprocessing involving class balancing using SMOTE, ensemble feature selection, and dimensionality reduction by truncated SVD. The Bi-LSTM model is fine-tuned by the Hybrid Hiking with Dollmaker Optimization Algorithm (HDOA) to forecast Urban Sustainability Scores (0–1). SHAP is employed for transparent visual explanation of feature contributions and decision paths. Along with better accuracy in predictions, the proposed method identifies prominent drivers of urban growth in green cover, population density, and connectivity through roads. Comparative analysis with recent hybrid models of literature brings out the robustness, explainability, and policy-relevance of our model. The insights drawn from this study provide scalable and actionable intelligence for urban planners and sustainability experts.