Intelligent Rental Valuation: A Full-Stack Approach to Multi-Modal Property Price Prediction Using Machine Learning and Deep Learning

The rapid expansion of digital property rental platforms has intensified the challenge of determining fair and competitive rental prices in highly dynamic markets. Property rental prices are influenced by a combination of structural attributes, geographical location, rental duration, and latent market demand, making manual pricing strategies unreliable and inconsistent. This research presents an intelligent, full-stack rental valuation system that integrates Machine Learning (ML) and Deep Learning (DL) models into a real-world property rental platform to deliver accurate, transparent, and scalable price predictions. The proposed system introduces a three-tier price prediction framework: manual feature-based prediction, where users provide explicit property attributes; geospatial map-based prediction, leveraging Google Maps integration to extract latitude and longitude for location-aware pricing; and CSV-based batch prediction, enabling bulk rental valuation for large property datasets. Ensemble regression models including Random Forest and XGBoost are evaluated alongside a deep neural network architecture to capture nonlinear relationships between property features and rental prices across daily, weekly, monthly, and yearly time horizons. Experimental evaluation using real-world rental datasets demonstrates that ensemble ML models achieve strong predictive performance in moderate data regimes, while DL models offer improved generalization as data volume increases. The bestperforming model achieved an R² score exceeding 0.90 with low Mean Absolute Error (MAE). The results validate the feasibility of deploying ML/DL-driven pricing intelligence within a live full-stack environment, contributing to fair pricing, reduced information asymmetry, and improved decision-making for both hosts and guests