THIRUVANANTHAPURAM: Soon, Kerala Police will be able to map accident hotspots across the state. In an effort to reduce rising accidents, the State Crime Records Bureau (SCRB) will launch an AI-powered road accident analytics and prediction suite to predict severity of road accidents and identify accident hotspots.
The work was started by January and the final model is expected to be launched by the end of the year. Sources told TNIE that data from 2018 onward has been collated and the AI models are being trained on that data. Parallel to this process, the SCRB has also initiated measures to ensure the quality of the accident data by personally verifying data from all the police stations in the state.
“Our team is validating data to enhance its quality. That’s a key part of the project,” said a source. The system will be able to predict the severity of accidents as fatal, grievous and minor based on parameters such as time, location, road and weather conditions, traffic density and historical data.
Another aim is to identify high-risk accident zones using spatial clustering and distance-based algorithms. This will enable decision-makers to come up with accident mitigation plans and launch measures such as enhanced patrolling of such zones to curb accidents. Another aspect is casualty prediction.
The probable number of casualties of road accidents can be predicted based on vehicle type, collision patterns, speed and environmental factors. The suite will also be able to deliver accident forecast on a monthly-basis at the state, district or even police station levels using historical trends and seasonality. For example, road accidents increase seasonally like during monsoon and this can be identified using the system and effective countermeasures can be drafted.
The system will also have features enabling it to keep a separate count on collision types such as head-on, rear-end, side-impact, rollover etc and that will give an idea about the causes of the accidents. Road accidents have been a cause of worry as 25,828 lives were lost on roads in the state from 2020 to 2025 November. During this period, there were 2.96 lakh road accidents and 3.3 lakh people sustained injuries.
Other objectives
Identify high-risk accident zones using spatial clustering and distance-based algorithms
Casualty prediction based on vehicle type, collision patterns, speed and environmental factors
Keep a separate count on collision types such as head-on, rear-end, side-impact, rollover to better understand causes of accidents