Dataiku > Case Studies > Real-Time Predictions for Targeted Safety Oversight

Real-Time Predictions for Targeted Safety Oversight

Customer Company Size

Large Corporate

Region

America

Country

Canada

Product

Dataiku

Tech Stack

Machine Learning
Data Science
A/B Testing

Implementation Scale

Enterprise-wide Deployment

Impact Metrics

Productivity Improvements
Cost Savings

Technology Category

Analytics & Modeling - Machine Learning
Analytics & Modeling - Predictive Analytics

Applicable Industries

Security & Public Safety

Applicable Functions

Business Operation

Use Cases

Predictive Maintenance

Services

Data Science Services

About The Customer

Technical Safety BC is an independent, self-funded organization mandated to oversee the safe installation and operation of technical systems and equipment across British Columbia (BC). In addition to issuing permits, licenses, and certificates, they work with industry to reduce safety risks through assessment, education and outreach, enforcement, and research. They were looking for a way to more accurately predict hazards and find more high-hazard sites while operating at the same resourcing level by introducing more sophisticated machine autonomy in the risk assessment process.

The Challenge

Technical Safety BC, an independent, self-funded organization, oversees the safe installation and operation of technical systems and equipment across British Columbia, Canada. Conducting physical assessments is costly, and false positive inspections can result in significant opportunity costs each year. Those same resources could be better allocated within the safety system; therefore, finding a way to more accurately predict hazards is of high strategic value to the organization, and it creates greater safety benefit to the public. Technical Safety BC was looking to find more high-hazard sites while operating at the same resourcing level by introducing more sophisticated machine autonomy in the risk assessment process. Some of the challenges faced included: uncoordinated heterogeneous data sources; data quality; speed of collaboration; and training challenges in the use of machine-recommended predictions.

The Solution

To address these challenges, Technical Safety BC created a dedicated data analytics and decision science department responsible for integrating advanced analytics into all parts of the organization. The team chose Dataiku as the tool to bring efficiency gains to the data process. Using Dataiku, Technical Safety BC is able to quickly prototype, test, iterate on, and deploy data-driven solutions, easily reuse models over and over again that work well rather than writing separate queries for similar projects, spend more time on innovative new ways to experiment with various models and work more efficiently, build a culture of experimentation and rapid prototyping, and accomplish more for safety outcomes with a small team of data scientists. Technical Safety BC used Dataiku on a project that leverages machine learning predictions to find common features and signals relating to risk factors. The predictive model adapts quickly to reflect any emerging risks and automatically shifts resource allocations accordingly based on the latest knowledge. A/B testing was used to test the new model that predicts high-hazard sites.

Operational Impact

Since deployment of the new machine learning model developed with Dataiku, Technical Safety BC’s predictive performance improved by 85% for the electrical technology compared to previous methods.
Machine learning also significantly reduced the total number of mandatory inspections, which will help optimize safety officer time so they can apply their time and expertise to potentially higher-hazard situations and make a greater impact on safety.
This new approach offers improved performance, shorter deployment time, better scalability, and paves the road for future improvements in service to public safety.

Quantitative Benefit

Predictive performance improved by 85% for the electrical technology compared to previous methods.
Significantly reduced the total number of mandatory inspections.
Improved performance, shorter deployment time, better scalability.