Show simple item record

dc.contributor.advisorMackaness, William
dc.contributor.authorWu, Xuchao
dc.date.accessioned2012-08-23T09:05:07Z
dc.date.available2012-08-23T09:05:07Z
dc.date.issued29/11/2012
dc.identifier.urihttp://hdl.handle.net/1842/6356
dc.description.abstractA large volume of work has been done to summarise the pattern of a pedestrian’s trajectory and to detect the intention of the pedestrian, based on their previously recorded trajectory. However, little effort has been devoted to implementing the real-time trajectory pattern prediction. This study is an innovative experiment which attempts to develop a system, employing machine learning algorithms to categorise the transportation mode and activity on the fly given the geospatial information from a pedestrian’s smartphone. Firstly, a Gaussian Mixture Model, aimed at predicting the transportation mode as well as a Naive Bayes Model and a Decision Tree Model, both of which aiming to predict a pedestrian’s current activity, are built, using two sets of training data. Then raw data collected from the smartphone is pre-processed and transformed into an ARFF format file. Finally, every record in the ARFF format file, which represents the instantaneous status of the pedestrian, is imported into the model and the result tables, which depict the probability of every transportation mode and activity of the pedestrian’s current status, are shown in the system.en
dc.language.isoenen
dc.publisherThe University of Edinburghen
dc.subjectPedestrian trajectoryen
dc.subjectReal-time predictionen
dc.subjectMachine learningen
dc.subjectGaussian Mixture Modelen
dc.subjectNaive Bayes Modelen
dc.subjectDecision Tree Modelen
dc.subjectMSc Geographical Information Scienceen
dc.subjectGISen
dc.titleSpace Book Project -- Trajectory Analysisen
dc.typeThesis or Dissertationen
dc.type.qualificationlevelMastersen
dc.type.qualificationnameMSc Master of Scienceen
dcterms.accessRightsRestricted Accessen_US


Files in this item

This item appears in the following Collection(s)

Show simple item record