Deep Learning and Hybrid Time-Series Forecasting for Intelligent Traffic Management and Sustainable Urban Mobility

Abstract: Urban traffic congestion is a major barrier to sustainable mobility because it increases travel delay, fuel consumption, emissions, crash exposure, and uncertainty in daily travel. Recent advances in connected vehicles, Internet of Things (IoT) infrastructure, urban sensing, navigation platforms, and deep learning have created new opportunities for intelligent traffic management systems that can forecast congestion before it spreads and respond through signal control, route guidance, and demand management. This paper presents a research-oriented review and conceptual framework for deep learning and hybrid time-series forecasting in intelligent traffic management and sustainable urban mobility. It uses the supplied reference papers on urban traffic management, connected and automated vehicle signal control, smart-city traffic optimization, anomaly-aware deep learning, hybrid CNN-LSTM, CNN-LSTM-GRU, GSA-LSTM, SDLSTM-ARIMA, attention-based GRU-LSTM, and CPO-CNN-LSTM-Attention models. The paper argues that future traffic systems should combine temporal learning, spatial network representation, anomaly detection, contextual data such as weather and holidays, optimization algorithms, and sustainability evaluation. For better presentation, the paper also includes comparative tables and conceptual figures that summarize model categories, architecture flow, and sustainability links.

Keywords: Deep Learning; Hybrid Forecasting; Time-Series Prediction; Intelligent Traffic Management; Sustainable Urban Mobility.