Incorporating twitter-based human activity information in spatial analysis of crashes in urban areas

The primary objective of this study was to investigate how to incorporate human activity information in spatial analysis of crashes in urban areas using Twitter check-in data. This study used the data collected from the City of Los Angeles in the United States to illustrate the procedure. The following five types of data were collected: crash data, human activity data, traditional traffic exposure variables, road network attributes and social-demographic data. A web crawler by Python was developed to collect the venue type information from the Twitter check-in data automatically. The human activities were classified into seven categories by the obtained venue types. The collected data were aggregated into 896 Traffic Analysis Zones (TAZ). Geographically weighted regression (GWR) models were developed to establish a relationship between the crash counts reported in a TAZ and various contributing factors. Comparative analyses were conducted to compare the performance of GWR models which considered traditional traffic exposure variables only, Twitter-based human activity variables only, and both traditional traffic exposure and Twitter-based human activity variables. The model specification results suggested that human activity variables significantly affected the crash counts in a TAZ. The results of comparative analyses suggested that the models which considered both traditional traffic exposure and human activity variables had the best goodness-of-fit in terms of the highest R2 and lowest AICc values. The finding seems to confirm the benefits of incorporating human activity information in spatial analysis of crashes using Twitter check-in data.