Evaluation of Machine Learning Algorithms for the Prediction of Simulated Company Parking Space Occupancy

Parking lots of cities and companies are congested when employees get to work in the morning hours. Some people might need to switch to public parking spaces if there is no spot left at their workplace. This is a serious challenge, as nowadays and especially in the future, the ratio of electric vehicles in transportation is growing. Therefore, electric vehicle users require guaranteed charging spots for the continuation of their journey. To enable reservations of parking and charging spots ahead of time, companies want to gain knowledge on upcoming vehicle numbers. So, there is a need for prediction models. However, there is little research done on prediction technologies for company parking spaces. In this paper, four different models from machine learning and statistics are compared in predicting the presence of vehicles from the user groups employees, fleet and visitors. They are trained on simulated occupancy data throughout one year. An evaluation shows that decision trees and artificial neural networks perform well for the case of company parking.