There has been a proliferation of bike share services around the world in the past few decades. Bike share programs are typically characterized by their convenience and efficiency. Ostensible benefits of bike share programs include less car dependency, fewer vehicle emissions, the ability to better link into existing public transit infrastructure, and potentially addressing the problem of first/last mile connectivity. Due to the data limitations, fewer studies have explored the temporal travel patterns of bike sharing program within a city and how they are related spatially. Using the extensively collected bike sharing log data of b-cycle service in Austin, San Antonio and Houston, this research addresses the temporal and spatial patterns of different kiosks in 2014 to 2016.
In this research, we want to focus on two problems in mining the data. A) What is the difference of the travel patterns of different membership group? B) What is the relationship of different temporal patterns of bike kiosks and there responding geolocations? And how can these patterns help contribute to the maintenance of the service?
In the research, we first try and conduct some basic analysis of data, including distribution of trip time, travel distance and their correlation with other properties like membership, trip type. Next, we try to use tools like Google Map and ArcGIS to visualize the geolocation of different kiosks, find the most frequently used kiosks, routes and spots of higher round trip ratios. Finally, we generate the temporal patterns of different kiosks as time series. Investigating for different learning algorithm that can help us cluster different patterns. Then we run the algorithm and plot the results on the map to observe the difference of those patterns and their geographical relations.