Can Machine Learning Pinpoint the Root of the Diabetes Epidemic?

A colleague of mine recently asked me to watch the Netflix documentary What the Health? It puts forth a narrative placing the blame of rising obesity and diabetes rates squarely on the shoulders of the American diet. In particular, the filmmaker takes great pains to target processed meats as a main culprit in what is undeniably a growing national epidemic.

After looking over the published literature myself I can see that the filmmaker took some significant liberties in his interpretation of various studies. His conclusions about the studies coming out of WHO are frankly unsupported. More importantly, many studies note the challenges of “residual confounding” when drawing conclusions. One example: high consumers of processed meats likely also drink milk and alcohol. Which one is to blame for increased obesity and subsequent diabetes?

My forte resides with politics, sales, business and other areas using demographic profiles to find interesting correlations via Artificial Intelligence and Machine Learning. Whether I’m looking for a genetic ancestor or a like-minded potential customer — using characteristics of populations to predict outcomes is always exciting. The same techniques can be used here.

A few months back the prolific John Battelle and I had an interesting exchange about this…