Hassan, BilalAlmansour, Rakan2024-03-312024-03-312024-03-09https://hdl.handle.net/20.500.14154/71734This research project is dedicated to advancing traffic recognition methodologies through the analysis of GPS-based data, aiming to address the persistent challenge of traffic congestion in urban and residential areas, especially during peak hours. The primary objective is to distinguish commuting behaviors among urban residents in their choice of public or private transportation modes. This distinction holds paramount importance in encouraging public transit utilization, potentially leading to time and cost savings for commuters. The project relies on GPS-based data, meticulously cataloged in Table 1, which undergoes comprehensive pre-processing to ensure compatibility with advanced machine learning techniques. Subsequently, an algorithm rooted in Artificial Neural Networks is developed to effectively differentiate between the two fundamental transportation modes: public and private. Leveraging factors such as image recognition and vehicle speed, the algorithm provides crucial insights into commuter behavior based on GPS data. The overarching objective is to offer predictive insights, empowering commuters to make informed choices between private and public transportation dynamically. Utilizing sophisticated Global Positioning System (GPS) technology, originally developed for military purposes and now extensively applied in various domestic scenarios, the project bridges the gap between intricate urban traffic patterns and state-of-the-art machine learning techniques. The envisioned outcome is not only an enhanced commuting experience but also a substantial reduction in traffic congestion during peak hours, contributing to an efficient and sustainable urban transportation ecosystem. It is noteworthy that the project specifically utilized GPS-based traffic data from a busy junction, publicly available for analysis. Employing the techniques discussed earlier, including the Recurrent Neural Networks (RNN), and Long Short-Term Memory (LSTM) algorithms, yielded promising results. The proposed Federated LSTM approach (FedLSTM) approach gave significant results where over the state-of-the-art approaches. These results contribute to the overall success of the project and reinforce its potential impact on improving urban transportation dynamics using GPS-based insights.58enMachine LearningArtificial IntelligenceTransportationTrafficNeural NetwrokThesis