Examination of the Effects of the Pandemic Process on the E-scooter Usage Behaviours of Individuals with Machine Learning

Abstract

Analysing user behaviour and thus travel mode choice is an important task in transport planning and policy-making in order to understand and predict travel demand. Recent pandemic events have challenged the modal choices of European users by reducing the use of public transport at various times and favouring walking and/or the use of electric bikes and scooters for last-mile travel. A number of studies have focused on analysing how the pandemic affected workers' choice of transport mode, with particular reference to local public transport, by developing multinomial logistic regression and artificial neural network models to analyse travellers' choice of transport mode before and after COVID-19. Particularly in non-European contexts, studies have been conducted on the relationship between socio-economic factors and the duration of e-scooter trips before and during the health crisis caused by the pandemic. in these contexts, a general increase in the duration of e-scooter trips after the pandemic was shown. Few studies, however, have analysed the European context. Several factors relating to services and infrastructure as well as socio-demographic components contributed to the propensity to use e-scooters as evidenced by a number of literature works in the European context. However, little research has been conducted using the machine learning approach to understand which factors and how they may influence modal choices. The present research work focused on the analysis of last-mile transport choices by investigating the propensity of a sample of users residing in Sicily during different time phases before and after the COVID-19 pandemic. In this study, 35 different classes were determined for a total of 545 data. The classification process was carried out using SMO, KNN and RF machine learning algorithms. The results showed a reduction in the frequency of e-scooter use during the health crisis caused by the pandemic. The results showed that this was a temporary behaviour, even though the purpose of e-scooter use by most individuals changed during the health crisis caused by the pandemic. However, it was observed that the frequency of e-scooter use decreased in most individuals during the health crisis caused by the pandemic and this became a permanent behaviour. The results suggest that the analysis of the importance of variables in relation to different periods and is essential for a better understanding and effective modelling of people's travel behaviour and for improving the attractiveness of these means of transport for companies operating services in the areas examined.

Keywords

• e-scooter
• machine learning
• pre-post covid-19 travel behaviour
• mobility choices
• sicily