#Predicting Bike Sharing Demand

INTRODUCTION Bike sharing systems have recently been adopted as a new means of transportation by a growing number of cities. These systems have not only automated the process of membership, rental and return; but have also provided a flexible, fast and green alternative for mobility. Users are able to rent a bike from a particular position and return back at another position. This increased flexibility poses the challenge of unpredictable demand as well as irregular flow pattern of the bikes. If we can predict the demand of bikes beforehand, we can prevent imbalance problems like unavailability of bikes or parking docks at the station. In this project, we aim to build such a model for predicting and classifying the number of bike-cycles given a set of input features.

DATA We obtained our dataset from the UCI Machine Learning Repository [1]. The data is related to the two-year historical log corresponding to years 2011 and 2012 from Capital Bike share system, Washington D.C., USA[2], along with the corresponding seasonal and weather information[3]. It consists of 17379 instances and 16 attributes- record index, date, season, year, month, hour, holiday, weekday, working day, weather, humidity, wind speed, temperature, count of casual users, count of registered users and the total count of users. The values of all the features except the ones to be predicted are normalized.