TITLE:
Demand Prediction of Ride-Hailing Pick-Up Location Using Ensemble Learning Methods
AUTHORS:
Divine Carson-Bell, Mawutor Adadevoh-Beckley, Kendra Kaitoo
KEYWORDS:
Ride-Hailing, Braess Paradox, Vehicle Clustering, Deadheading, Congestion, Predictive Modelling, Vehicle Deployment, Ensemble Learning
JOURNAL NAME:
Journal of Transportation Technologies,
Vol.11 No.2,
April
25,
2021
ABSTRACT: Ride-hailing and carpooling platforms have become a popular way to move
around in urban cities. Based on the principle of matching riders with drivers,
with Uber, Lyft and Didi having the largest market share. The challenge remains being able to optimally match rider demand
with driver supply, reducing congestion and emissions associated with Vehicle
clustering, deadheading, ultimately leading to surge pricing where
providers raise the price of the trip in order to attract drivers into such
zones. This sudden spike in rates is seen by many riders as disincentive on the
service provided. In this paper, data mining techniques are applied to
ultimately develop an ensemble learning model based on historical data from
City of Chicago Transport provider’s dataset. The objective is to develop a
dynamic model capable of predicting rider drop-off location using pick-up
location data then subsequently using drop-off
location data to predict pick-up points for effective driver deployment under multiple scenarios of privacy and
information. Results show neural network algorithms perform best in
generalizing pick-up and drop-off points when
given only starting point information. Ensemble learning methods, Adaboost and Random forest algorithm are able to predict both drop-off and
pick-up points with a MAE of one (1) community area knowing rider pick-up point and Census Tract information only and in
reverse predict potential pick-up points using the Drop-off point as the
new starting point.