Understanding travel behaviour and travel demand is of constant importance to transportation communities and agencies in every country. Nowadays, attempts have been made to automatically infer the modes of transport from positional data (such as GPS data) to significantly reduce the cost in time and budget of conventional travel diary surveys. Some limitations, however, exist in the literature, in aspects of data collection (spatio-temporal sample distribution, duration of study, granularity of data, device type), data pre-processing (managing GPS errors, choice of modes, trip information generalisation, data labelling strategy), the classification method used and the choice of variables used for classification, track segmentation methods used (clustering techniques), and using transport network datasets. Therefore, this research attempts to fully understand these aspects and their effect on the process of inference of mode of transport. Furthermore, this research aims to solve a classification problem of sparse GPS data into different transportation modes (car, walk, cycle, underground, train and bus). To address the data collection issues, we conduct studies that aim to identify a representative sample distribution, study duration, and data collection rate that best suits the purpose of this study. As for the data pre-processing issues, we standardise guidelines for managing GPS errors and the required level of detail of the collected trip information. We also develop an online WebGIS-based travel diary that allows users to view, edit, and validate their track information to assure obtaining high quality information. After addressing the validation issues, we develop an inference framework to detect the mode of transport from the collected data. We first study the variables that could contribute positively to this classification, and statistically quantify their discriminatory power using ANOVA analysis. We then introduce a novel approach to carry out this inference using a framework based on Support Vector Machines (SVMs) classification. The classification process is followed by a segmentation phase that identifies stops, change points and indoor activity in GPS tracks using an innovative trajectory clustering technique developed for this purpose. The final phase of the framework develops a network matching technique that verifies the classification and segmentation results by testing their obedience to rules and restrictions of different transport networks. The framework is tested using coarse-grained GPS data, which has been avoided in previous studies, achieving almost 90% accuracy with a Kappa statistic reflecting almost perfect agreement.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:632026 |
| Date | January 2014 |
| Creators | Bolbol, A. S. Z. |
| Publisher | University College London (University of London) |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://discovery.ucl.ac.uk/1448075/ |
Page generated in 0.0018 seconds