Demand Modelling Based on Geostatic Data Calibrated Using Mobile Network Data
Information on how people move within urban areas enables a wide range of application. Shared mobility companies can use this information to optimize their business areas, public transportation companies can enhance their timetables and routes. Recently, mobile network providers have started distributing their data as origin-destination matrices as a basis for such use-cases. As part of our research, we have developed a demand model trained on aggregated mobile network data to address areas not covered by the provider’s data. A model is presented that uses correlations between movements and geostatical data such as population or land usage data. Multiple machine learning approaches were trained and tested, including methods like LASSO, Kernel-Ridge, Gardient Booosting and ANNs. The final model enables predictions for any (urban) area in Europe.