In the project P2Map I am trying to combine objective sensor data with subjective context information and to use this combined information in machine learning methods to predict a geospatial and temporal distribution of air quality.
Therefore my interests spread across data mining and machine learning methods in general and more specificly methods for spatiotemporal data, for example land use regression. I am trying to develop a more general approach, because most land use regression models are designed for certain regions and rely on land usage data that may be only available in this region. For this task, neural networks, especially convolutional neural networks, are tested at the moment.
In another facet of the P2Map-project, an array of low cost sensors has to be calibrated in order to reliably measure the surrounding air quality. Finally for the incorporation of the subjective data, like context information or perceptions, I will be working with techniques from the field of natural language processing.
Introductions to algorithms and data structures:
Seminar "Web 2.0":
Seminar "Machine Learning":
MapLUR: Exploring a New Paradigm for Estimating Air Pollution Using Deep Learning on Map Images in ACM Trans. Spatial Algorithms Syst. (2020). 6(3)
OpenLUR: Off-the-shelf air pollution modeling with open features and machine learning in Atmospheric Environment (2020). 233 117535.
SimLoss: Class Similarities in Cross Entropy (2020).
Anomaly Detection in Beehives using Deep Recurrent Autoencoders in Proceedings of the 9th International Conference on Sensor Networks (SENSORNETS 2020) (2020). 142–149.
EveryAware Gears: A Tool to visualize and analyze all types of Citizen Science Data in Proceedings of VGI Geovisual Analytics Workshop, colocated with BDVA 2018, D. Burghardt, S. Chen, G. Andrienko, N. Andrienko, R. Purves, A. Diehl (reds.) (2018).
Air Trails--Urban Air Quality Campaign Exploration Patterns in AGILE Workshop (2018).