Max Rerisi

Type 2 diabetes affects hundreds of millions of people worldwide, many of whom do not even know they have the disease. Diabetes arises from the body's inability to regulate blood sugar levels. If gone unnoticed and untreated, diabetes can cause severe, life-threatening symptoms. Changes like healthier diets or medication are necessary for people with diabetes to prevent fatal complications. Any technique that can provide further insight into the disease and diagnose individuals sooner can improve survival rates, and breakthroughs in the field can keep the number of new type 2 diabetics to a minimum. Previous studies have used machine learning (ML) to see how different amounts and thoroughness of data can improve performance. ML is a process that involves taking large amounts of data and training a mathematical algorithm to make predictions about new data by feeding it subsets of data. 

In my project, I am attempting to provide insight into diabetes through these models. While they can be used to predict diabetes to varying degrees of accuracy, what I hope to do is apply a more useful technique that measures the weight, or importance, of each attribute the datasets have. The more important the attribute, the more related it is, directly or indirectly, to a patient’s current condition or why they do (or don’t) have diabetes. The combination of attributes with the source of the data can be used to conclude what causes the attributes to change and how or why they impact diabetes. This research and consideration can help protect against new cases of diabetes and can improve the condition of current diabetics as new insight is gained into what can cause negative symptoms. A particular example this can be applied to is the diet of Eastern vs. Western cultures. Determining fundamental nutritional differences in typical Eastern/Western diets and their effect on diabetes is just one such case in these techniques, and this field can be utilized by comparing the differences between the types of patients and their medical information.