Classification Driver's Behaviour Using Supervised Algorithm

At the present time, we discuss the human behavior of driving and death rates due to an accident on the road around the world. Hence, the real-time response of notification about the risk on road is insufficient. Moreover, the most problem is people's lack of knowledge for driving, especially people careless while driving that may lead to an accident. Driver's behavior classification is required in order to prevent unfortunate accidents on the road. Many previous studies, researchers focused on simulation driver and limited road pattern to collect data for classification. However, the main problem is the data is inadequate and the driver's data should be collected from the driver's daily life to get an effective classification. This work deals with an efficient supervised learning procedure to predict driver's behavior by comparison from five classifiers and vote the highest score to predict data. All data are collected from sensors embedded in the vehicle's in Indonesia. Throughout the dataset over one million records, DBC which classify Aggressive and Non-aggressive, the result show F1-score is 86% of twenty thousand labels.