Fire Safety Code Inspections: A predictive fire inspection AI solution

The Problem

The number of scheduled proactive fire safety inspections have doubled over 8 years while request-based inspections have grown exponentially, creating a critical systemic imbalance. To restore sustainability, the City must implement data-driven prioritization and new reporting tools to focus expert resources on high-impact initiatives and time-sensitive public safety needs.

The Solution

To manage unsustainable inspection volumes, the City implemented a machine learning model that prioritizes fire safety compliance inspections by focusing on properties most likely to fail. The model leverages data inputs, such as occupancy type, past inspection failure rates and the time elapsed since the last inspection to predict high-risk cases. This data-driven strategy successfully captures approximately 90% of failures while enabling more efficient resource allocation. The model was built using open source tools within existing budget.

The Impact

The project applies to approximately 15,000 properties inspected annually or bi-annually across the city. Optimized resource allocation by significantly reducing the total number of inspections needed.

Lessons Learned

It is essential to continue inspecting some low-probability properties to gather data for ongoing model training and sustainability.

Launch year: 2017

Alberta Safety Codes Act, Alberta Fire Code, Quality Management Plan (QMP).