This paper addresses the critical need for accurate traffic flow forecasting in urban areas to mitigate traffic congestion, environmental pollution, and safety risks. The authors propose a two-level machine learning approach that combines unsupervised clustering and supervised machine learning models. The first level uses an unsupervised clustering model to extract patterns from sensor-generated data, while the second level employs supervised machine learning models to predict traffic flow within each cluster. The approach is evaluated using real data from traffic sensors in Catania, Italy, and is tested on both working days and holidays to validate its effectiveness. The study highlights the potential of the proposed method in enhancing traffic management and urban planning by providing accurate and timely traffic flow predictions. The results show that the XGBoost algorithm outperforms other models in terms of accuracy, making it the preferred choice for short-term forecasting. The approach demonstrates the ability to handle sparse data and predict traffic flow for roads with limited sensor coverage, contributing to more efficient and sustainable urban transportation systems.This paper addresses the critical need for accurate traffic flow forecasting in urban areas to mitigate traffic congestion, environmental pollution, and safety risks. The authors propose a two-level machine learning approach that combines unsupervised clustering and supervised machine learning models. The first level uses an unsupervised clustering model to extract patterns from sensor-generated data, while the second level employs supervised machine learning models to predict traffic flow within each cluster. The approach is evaluated using real data from traffic sensors in Catania, Italy, and is tested on both working days and holidays to validate its effectiveness. The study highlights the potential of the proposed method in enhancing traffic management and urban planning by providing accurate and timely traffic flow predictions. The results show that the XGBoost algorithm outperforms other models in terms of accuracy, making it the preferred choice for short-term forecasting. The approach demonstrates the ability to handle sparse data and predict traffic flow for roads with limited sensor coverage, contributing to more efficient and sustainable urban transportation systems.