A satellite remote sensing technique, Interferometric Synthetic Aperture Radar (InSAR), is able to provide surface displacement information on a millimeter level. In this study, data from the TerraSAR-X satellite collected in the years 2009-2018 over the area of Amsterdam is used. Even though radar data is a subject to multi-step processing, there are still several problems observed that can make the interpretation of the time-series difficult for users who are not experts in the radar remote sensing field. In this study we focus on unwrapping errors, partial decorrelation, and incorrectly fitted models. The unwrapping errors are handled as outlier detection problem and the rest as a time-series segmentation task. In order to address these issues, a data-driven approach is proposed. We show a method to detect unwrapping errors based on spatially neighboring points. A GUI is developed to collect expert knowledge in a form of assessing the time-series correctness, marking unwrapping errors, and dividing time-series into separate segments. This information is later used in the evaluation of several outlier detection and segmentation algorithms. We propose a supervised learning approach based on neural networks in order to use expert knowledge. Due to not enough labelled data available, a simulation is developed and used for training of the networks. We present two different approaches, one based on multi-label classification and one on binary classification. For each of them fully-connected neural networks and convolutional neural networks are compared.