A Swine Health Information Center-funded project is facilitating the expansion of farm-level outbreak forecasting to new regions.
The SHIC plans to expand a project that uses machine learning to calculate when a sow farm will break with Porcine Epidemic Diarrhea.
Researchers at the University of Minnesota built machine learning algorithms that consider factors such as recent animal movements, current disease distribution and environmental factors to predict whether a sow farm will break with PED two weeks in advance.
Swine Health Information Center Executive Director Dr. Paul Sundberg said in contrast, the swine sector better understands the epidemiology of PED virus and Porcine Reproductive and Respiratory Syndrome virus. The ability of producers to effectively estimate the risk of disease outbreak risk is lacking.
Dr. Sundberg said the industry could predict an outbreak of PED on a farm two weeks before it would happen; think about the opportunity to ramp up biosecurity or ramp-up management. Or maybe early weaning or other things that could happen on that farm to prevent or proactively respond to a predicted outbreak.
The issue here is not to cry wolf and say, oh gosh, something’s going to happen, and then it never does. The point is to get this whole program as effective, as targeted, and as efficient as possible to accurately predict an outbreak of PED with the confidence it will happen.
“We’re making good steps toward that goal and I think we’re almost there and when we do, I’m looking forward to applying it to other diseases. We may be able to apply it to PRRS, for example, which would be a major step in PRRS control and PRRS management.”
Dr. Sundberg said the farm-level outbreak forecast results from producers willing to openly share their information to analyze the data better and then feedback to their benefit. One of the objectives and the mission statement of the SHIC is for the analysis of data. The Centre is thriving in doing some computer learning projects in Southeast U.S. and taking into account the topography around farms and pig movements, such as weather patterns, disease situation, and status around farms in the Southeast U.S.
“Put it into an algorithm for a computer to learn when to predict a farm would break with the disease.”
The exciting results the Centre achieved in the Southeast U.S. encouraged them to expand into the U.S. Great Plains. Different topography, weather patterns, and the same issues of pig movement and disease status around the farm still exist.
“But the sharing of the information, I think is the thing to highlight here, and the ability to use that to the betterment of herd health and to predict diseases before they happen.”
He said the more factors they can enter into the machine learning program, the more accurate it is.
These projects reports are up on the swinehealth.org website and a sign up for a monthly newsletter that includes project updates. It consists of information on a specific lay on this project and others that we have going on, as well as the domestic and international disease monitoring reports that come out once a month.
Dr. Sundberg said the opportunity to take advantage of producers voluntarily sharing their information is exciting.
“It’s a good example of the strength in numbers and how it strengthens the industry when voluntarily talking to each other, rather than stay in our silos.”
He thinks it is an excellent opportunity to highlight the progress of that cooperation to the benefit of the whole pork industry. •
— By Harry Siemens
