Company
BACKGROUND
The client has access to 433k chargers throughout Europe. We focus on the city of Berlin for a PoC.
Problem
It is not clear what are the success factors of an electric vehicle (EV) charger. Therefore, the proper location of new charging stations is unknown. Additionally, EV users can not schedule their charging due to the lack of ML.
Benefit
Among others, prediction of hourly charger utilization leads to: identification of best location for new chargers; charging scheduling for EV users; better city planning.
METHODOLOGY & results
Architecture: Local. Google Colab.
Developing language: Python
Data mining frameworks: CRISP-DM.
ML techniques: Attribute reduction whit K-Means
Supervised learning whit Decisions Tree Regressions and Random Forest.
Results: Accuracy between 82% and 97% depending on the hour of the day
