Abstract:
Talking on the phone or being distracted by events outside the car clearly can lead to the loss
of the driver’s alertness or vigilance.
To predict whether the driver is alert or not has therefore became a main issue for cars
Companies such as Ford. This research study is based on the prediction challenge proposed by
the company Ford via the website Kaggle.com. The objective is to be on the top 5 on the
Kaggle Leaderoard where participants are ranked according to the score their model achieved.
Many binary prediction model or binary classifiers can be used for such challenges. A further
investigation on the data shows that linear models are the most adapted.
Logistic regression with feature engineering have been performed on the data to come out
with a final classifier.
This model clearly achieve a better score than the winner of the competition and use only 4
variables.