EFFECT ASSESSMENT OF LARGE-SCALE EVENTS VIA SPATIOTEMPORAL APPROACH

Abstract. Together with rapid development of location-based services and big-data platforms especially in urban areas, huge amount of spatiotemporal data are collected without properly used; on the other hand, state-of-the-art quantitative policy effect assessment techniques usually require panel data as input. To solve both issues, this paper follows the following approach: obtaining panel data by aggregating spatiotemporal data and feeding them to the effect assessment module. With the help of high-performance computing techniques which are able to deal with huge amount of data, we build framework Aggr-analysis which applies clustering algorithms to shrink the raw data set and find associations between different data sets via co-location analysis. Finally, we prove the effectiveness by an example: analysis of resident activities during the COVID-19 Pandemic. We apply Aggr-analysis to process the share-bike usage data and POI (Point Of Interest) data in Beijing, then obtain the panel data required by DID (Difference-in-Differences) method. Supplemented with environmental data, we conclude the net effect of the COVID-19 breakout on society and economy - the pandemic has reduced the overall resident mobility by 64.8% within two months.