IDENTIFY ABNORMAL DRIVER BEHAVIOR THROUGH
DEEP LEARNING

According to the Association for Safe International Road Travel around 130,00,000 people lost their life on roads due to accidents. Road accidents claim one life every four minutes in India, most of which are cases of drunk driving. According to the transport research program conducted by the Indian Institute of Technology, Delhi, 141,526 persons lost their lives and around 477,731 got injured in the road accidents in India in 2014 alone. These numbers may just be an exaggeration when compared to the actual scenario, adding to the fact that most of the accidents go unrecorded. This throws light on the importance of having safe roads and more importantly safe drivers.
 
An analysis to communicate the emphasis of safety of drivers using advanced anomaly detection algorithm has been studied. The Internet of Things (IoT) which is a ubiquitous technology instituted the diagnosis of driving patterns that eventuated a collision. Real- time driving data was collected on an experimental basis from the OBD II port of a small segment car without interfering the driver. The erratic driving pattern analysis were distinguished using a five-layer neural network which was implemented through Python. The cross validation of the prototypes enhanced the discoveries stated in this paper.