Bike-ride sharing systems are the new generation of traditional bike rentals, where the entire process is automated. A user rents a bike from one location and returns it at another location. There are more than 500 bike-ride sharing systems around the world, consisting of more than 500,000 bikes. Bike-ride sharing systems are typically found in urban and large cities such as Boston, N.Y. City, Washington DC, Paris, Montreal, and Barcelona. Bike-ride sharing is particularly important due to their important impact on traffic, environment, and health. As popular as bike-ride sharing systems are, there is a lack of a reliable model to forecast (predict) bike rental demand daily. Lack of available bikes constitutes an inconvenience to individuals seeking a bike at a certain location and a loss of revenues for companies operating the bikes. This paper develops a Machine Learning (ML) model (algorithm) to forecast (predict) the number of bikes rented daily based on historical data. Moreover, the model overlays environmental and seasonal settings to study their impact on bike rental demand. We test our ML model using a real-life dataset obtained from a local bike-ride sharing company in the City of Boston in the state of Massachusetts in the United States. We also applied the model to historical dataset from New York City (NYC). In both cases, the model is accurate and reliable.
Read full abstract