Shared Bike Scheme in New York: Predicting the end of trips with Supervised Machine Learning

Sacristan, Pedro (2014) Shared Bike Scheme in New York: Predicting the end of trips with Supervised Machine Learning. Diploma thesis, Dublin, National College of Ireland.

PDF (Diploma)
Download (3MB) | Preview


Share bike schemes are increasingly reliant on technology to improve and expand its services (AAAI, 2009). Being able to develop efficient, reliable and sustainable bike scheme systems will facilitate the accommodation of the rapidly increasing population in most of the cities around the world (IEEE, 2013).

This paper examines the possibility of predicting the destination of a trip by bike after the subscriber starts the journey. For that purpose some machine learning techniques have been applied to a dataset which contains almost one year of bike trips in the Share Bike Scheme of New York City.

Two supervised machine learning algorithms developed in a package of R have been used. “Naïve Bayes” was the first one but, because of the poor performance, a second one was used too, “Gradient Boosting Machine“.

Due to the lack of investment and means to carry out a complete analysis over the whole dataset a simplification was done by studying the performance of both algorithms in a sub-dataset of nine stations. It has been proven that the accuracy of the model is larger than was expected and that having the anonymized data the accuracy could reach a higher value.

The main contribution that this project makes to the subject of bike urban mobility is the potential for more precision than the time average of past events in predicting times for a bike in an empty station.

Item Type: Thesis (Diploma)
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
Divisions: School of Computing > Higher Diploma in Science in Data Analytics
Date Deposited: 16 Dec 2014 14:20
Last Modified: 16 Dec 2014 14:20

Actions (login required)

View Item View Item