This paper presents a novel approach that utilizes machine learning techniques, specifically clustering algorithms and artificial neural networks, to improve the prediction and understanding of human routines in urban mobility contexts. Our method focuses on the identification and categorization of Points of Interest (POIs) from travel data, facilitating the accurate prediction of user routines for intelligent transportation systems. By integrating a clustering phase that groups individual stop points into POIs, followed by a correction mechanism through user interaction, we address the limitations of existing methods in adapting to dynamic mobility patterns and the contextual ambiguity of GPS coordinates. Subsequently, a classification phase employs a feed-forward neural network to assign incoming travel events to the identified POIs. This dual-phase approach not only improves the precision of routine predictions but also enhances the adaptability of the system to changes in mobility behavior over time. The incorporation of a cognitive module, based on Dynamic Neural Fields (DNF), further allows for personalized predictions regarding the timing, duration, and nature of trips. Validated with datasets from the Portuguese city of Braga, our results demonstrate the effectiveness of this methodology in providing actionable insights for the development of cognitive solutions for the project BE.Neutral's innovative vehicle "BEN". By emphasizing user involvement and algorithmic transparency, our work contributes to the advancement of smart transportation technologies in urban environments. |
*** Title, author list and abstract as seen in the Camera-Ready version of the paper that was provided to Conference Committee. Small changes that may have occurred during processing by Springer may not appear in this window.