bike-sharing-predictor

Bike sharing systems have been gaining prominence all over the world with more than 500 successful systems being deployed in major cities like New York, Washington, London. With an increasing awareness of the harms of fossil based mean of transportation, problems of traffic congestion in cities and increasing health consciousness in urban areas, citizens are adopting bike sharing systems with zest. Even developing countries like India are adopting the trend with a bike sharing system in the pipeline for Karnataka. This paper tackles the problem of predicting the number of bikes which will be rented at any given hour in a given city, henceforth referred to as the problem of ‘Bike Sharing Demand’. In this vein, this paper investigates the efficacy of standard machine learning techniques namely SVM, Regression, Random Forests, Boosting by implementing and analyzing their performance with respect to each other.This paper also presents two novel methods, Linear Combination and Discriminating Linear Combination, for the ‘Bike Sharing Demand’ problem which supersede the aforementioned techniques as good estimates in the real world.