Demand prediction for shared mobility services using time series modelling

Abstract

Users are becoming more interested in diverse mobility solutions; however, as promising as it sounds, this type of mobility has its pain points. With an effective technological assistance tool, shared mobility services will not only offer a better user experience but also help in reducing traffic congestions and other daily struggles in the transportation industry. The main objective of this thesis is to analyse and investigate the possibilities of optimising shared mobility using historical data by predicting the total number of generated requests per hour. With the use of a Maltese ride-hailing operator’s data, the study first investigates where and how pickup requests were made. Subsequently, the data is examined for any time-series patterns, primarily trend, seasonality and cyclic. Different types of parametric techniques like Holt- Winter, Autoregressive Integrated Moving Average (ARIMA) and Facebook Prophet were used to compare whether seasonality and other characteristics have any effect on the performance of the model. From the results obtained, it can be implied that exogenous or independent data such as temperature and public holidays do not affect the predictive model. Forecasting models are validated by splitting the data between a training set for model fitting and a testing set to compare the outcome with the observed values. Metric performance results conclude that amongst all models, the accuracy of Holt-Winter’s outperforms other models with an overall Mean Absolute Error (MAE) of 8.039 and Root Mean Squared Error (RMSE) of 11.159.