From individual trajectories to system flows: A multimodal mobility analysis framework using mobile phone application data

The decarbonisation of the transport sector necessitates a paradigm shift from car-centric planning to sustainable, multimodal mobility systems. However, effective monitoring of these dynamic networks remains challenging due to the limited spatial coverage of traditional static sensors and the dense, high-quality labelled data typically required by emerging deep learning approaches. This study introduces a robust, data-driven framework that leverages sparse, large-scale mobile phone application data to bridge the gap between individual behaviour analysis and system-level flow monitoring. We propose a hierarchical travel mode detection algorithm integrating Support Vector Machines (SVM) with GIS-based map matching, specifically optimised to handle the irregular sampling rates and positioning noise inherent in passive crowdsourced data. Beyond identifying individual modes with 89.56% accuracy, this research advances a bottom-up aggregation methodology to construct high-frequency, street-level multimodal traffic flows. Validated against ground-truth multimodal traffic sensors in London, our approach demonstrates robust spatial consistency and strong correlations ( r > 0.70) across daily and hourly temporal resolutions. These findings establish that passive mobile phone data can serve as a cost-effective, scalable proxy for continuous transport system monitoring, enabling granular analysis of daily traffic dynamics and network-wide performance that traditional models fail to capture.