Using machine learning, we can predict the severity, location, and timing of accidents to help first responders with resource allocation and safeguard new drivers. By being able to foresee the likelihood of collisions based on driving conditions along with the fatality, the government could leverage this information to strategically position/allocate first responders across the city. Aswell, oftentimes many new drivers are exposed to unfamiliar conditions in which they do not know how to maneuver accordingly. Nationwide, almost “43% of first-year drivers are involved in accidents” (Safety Insurance Company), which is a leading concern for both drivers and the well-being of pedestrians. Modeling this problem will allow us to target and pinpoint which conditions yield the greatest risk and hence we can better prepare new drivers in the form of revised driving examinations, regulations, and information manuals.

By being able to predict the likelihood, severity, and time of accidents in certain regions, we can help first responders (paramedics, firefighters, police) be in a better position to handle the influx of accidents. There has been an extreme shortage of first responders in Canada and the ministry has been struggling to divide and disperse the workforce strategically. Interestingly with COVID-19, the workforce had taken a hit on staff supply and financial aid. Given the conditions and availability of resources, if we can correlate and predict the severity and timing of collisions, assets could be efficiently allocated to reduce stress on first responders.

Boosted Decision Tree Results: Precision: 0.87 Recall: 0.75 F1: 0.81