As debates around the use of artificial intelligence and the corresponding need for AI ethics heat up, so does the world around us (literally).
Extreme heat is bad news across industries, from insurance and health care to construction and agriculture. But it’s individuals laboring outdoors who bear the brunt of rising global temperatures and increasingly frequent heat waves.
According to the BBC, over 67,000 people are hospitalized each year in the US alone due to heat. The US Bureau of Labor Statistics reported in 2021 that 436 work-related deaths due to heat exposure had occurred since 2011.
For the 2024 SAS Hackathon StaSASticians team, the challenge was clear.
How can we protect workers in real time, before heat illness strikes?
The problem of heat stroke in construction
Because of strenuous physical activity and prolonged exposure to the elements, construction workers are especially vulnerable to heat stroke and heat-related illnesses. But traditional methods of monitoring heat stress are reactive and based on visible symptoms.
In a setting like this, every second counts. Any delay in responding to the signs of heat illness can spike worker safety issues, related health risks and the likelihood of insurance claims.
The StaSASticians set out to develop a real-time solution that would collect relevant information to detect early signs of illness before harm is done. Their innovative solution holds promise for many workers and industries.
Best of all, it can be a lifesaver for construction workers facing the heat.
Protecting workers with AI and IoT
To develop their AI-enabled heat stroke prevention system, the StaSASticians used SAS® Viya®, SAS Viya Workbench, SAS® Data Maker and Python tools. The solution uses IoT-equipped smart helmets from BeeInventor (a team member) to collect real-time physiological data from workers, like heart rate and core body temperature. Other data from the environment, including humidity and ambient temperature, are also collected and sent to the system.
Through machine learning and GenAI techniques, the system integrates environmental and personal data, analyzes it and calculates a physiological strain index (PSI) for each worker.
As information is collected, the PSI is displayed on a dashboard that provides real-time warnings and recommendations for preventing heat stroke. Supervisors monitoring the dashboard can easily identify workers at risk and act quickly before symptoms progress to a critical state. The solution helps to:
- Save lives and enhance safety and productivity.
- Support regulatory compliance.
- Reduce health care costs.
- Reduce the likelihood and cost of insurance claims.
The heat stroke prevention system protects not just high-risk workers, but also construction firms that confront high health care and insurance claims costs. Because the system can integrate with existing construction industry safety protocols and technologies, firms have even more incentive to adopt it.
Techniques of the solution
Sophisticated tools work behind the scenes to put the heat stroke prevention solution into action. Some elements of the solution are:
- Data collection. Smart helmets continuously monitor workers’ physiological data and environmental conditions.
- Data analysis. Collected data is processed and analyzed using AI models developed on SAS Viya Workbench, Python and R.
- Real-time alerts. The system generates real-time alerts and recommendations to prevent heat stress and ensure timely interventions.
- Synthetic data generation. Using SAS Data Maker to generate synthetic data helped the team solve issues with the data to ensure robust model training and validation.
- Continuous improvement. The AI models are continuously updated and refined based on new data, improving their accuracy and effectiveness over time.
- Cloud computing. The system relies on cloud computing environments for big data preparation and automated AI model development, ensuring scalability and flexibility.
Why is synthetic data so valuable?
One of the challenges the team had when developing the heat stroke prevention system was imbalanced real-world data (in other words, not enough diverse data). Without sufficient training data, AI models are not reliable or accurate.
Using the SMOTE technique to supplement their real-world data, the team built robust AI models that could accurately predict and prevent heat stroke incidents. And they did it in just a few weeks.
Synthetic data for insurance
Synthetic data is particularly valuable in the insurance industry, where data privacy and regulatory compliance are crucial. With synthetic data, insurers can generate realistic data sets without exposing sensitive information, opening the door to more effective
Profound implications for insurance and beyond
Preventing heat strokes is a winner on many levels. As insurers face increasing claims and higher costs due to weather extremes, it’s a clear winner for them as well.
Heat stroke prevention and fewer heat illness-related incidents can significantly reduce the number of insurance claims, enabling insurers to offer discounts to construction firms. Construction firms that adopt the system could reduce the number of workers’ compensation claims while maintaining a higher level of workplace safety. And because the system’s real-time insights help construction firms assess and manage risks more effectively, it can lead to more accurate underwriting and pricing by their insurance carriers.
Not just for the lab
Hackathons are great opportunities to test out theories, try new techniques and win well-deserved recognition for meaningful work. The StaSASticians team, runner-up to the 2024 SAS Hackathon grand champion (Team Butterflies), was also a winner in three separate categories: Americas (region), Insurance (industry) and IoT (technology).
But Hackathons are not just theoretical or confined to labs. The StaSASticians team is planning some early field testing of its heat stroke prevention system later this summer in Asia.
