Extracting Mobility-Relevant Information from Crowdsourced GPS Data

Learning knowledge from users GPS traces can provide rich context information to be applied in several areas. However, without processing, extraction of meaning can be impractical or a time consuming activity. The data used was collected using SenseMyFEUP application and represents real data, as a research involving real data the first step is to clear the data of errors like outliers in position, speed and time. The main focus of this research isn't the data filtering but the treatment of crowdsourced data, for that an approach is proposed to reduce the GPS trace to meaningful aggregated data and automatically infer the transportation mode used in a trip. The approach consists of four parts: a change-point based segmentation method, a clustering algorithm, an inference model to deduce the transportation mode and a trip chaining algorithm to merge trips identified at first as one but that are more meaningful together