|dc.description.abstract||In response to increasing urbanization, China seeks alternative public transportation methods, such as bike- sharing, which has demonstrated social and environmental benefits. As a result, the number of bike-sharing programs has grown rapidly over the last five years in China. Our sponsoring company TalkingData collects bike-sharing usage data via smartphones and aims to provide insights to bike-sharing operators. Therefore, our main objective in this project is to analyze smartphone data to understand the bike-sharing demand in Beijing. We conducted interviews with bike-sharing stakeholders and investigated a one-month sample of data that TalkingData collected from bike-sharing operators in Beijing merging it with secondary data from online resources.
We found that the level of bike-sharing activity varies across the city Beijing both in terms of location and time throughout the day. We discovered that both, time and environmental related factors significantly affect the bike-sharing demand. In contrast, our study revealed that some factors stated in literature such as pollution level do not affect bike-sharing demand in Beijing significantly. Hence, we suggest that drivers of bike-sharing demand differ across cities or countries making it worthwhile to perform location specific analysis. We fitted linear regressions, neural networks and random forests on the compiled dataset and compared their respective performance. We found that, based on the one-month sample, linear regression performs best amongst the three models in predicting hourly bike-sharing demand in Beijing.||en_US