Quantifying Parking Difficulty with Transport and Prediction Models for Travel Mode Choice Modelling.

Promoting sustainable transportation necessitates understanding what makes people select individual travel modes. Hence, classifiers are trained to predict travel modes, such as the use of private cars vs bikes for individual journeys in the cities. In this work, we focus on parking-related factors to propose how survey data, including spatial data and origin-destination matrices of the transport model, can be transformed into features. Next, we propose how the impact of the newly proposed features on classifiers trained with different machine learning methods can be evaluated. Results of the extensive evaluation show that the features proposed in this study can significantly increase the accuracy of travel mode choice predictions.