Many data analysis tasks involve sampling and recording of data. One of the first steps towards obtaining a coherent analysis is outlier detection. Though, outliers are often considered as an error or noise, they sometimes carry important information. If outliers are not detected, it may lead to biased parameter estimation, incorrect results, and wrong model specification. It is therefore, essential to identify outliers before modelling and data analysis. Time series is one of the application areas of outlier detection. A typical example of outlier detection application in time series is outlier detection among GPS trajectories. This research will focus on this area and on improving the results of existing methods especially; GPS segment averaging result.

Supervisor: Prof. Pasi Fränti