Predicting Diabetes Mellitus Using Machine Learning and Optical Character Recognition

Abstract

Diabetes Mellitus is recognized as a chronic metabolic disease that is characterized by hyperglycemia. As stated by the International Diabetes Federation, the statistics reveal that the incidence of diabetes among adults in Sri Lanka is 8.5%. In hindsight, this indicates that an average of one in every twelve adults in Sri Lanka is at risk of being diagnosed with the disease. However, presently, due to the lack of knowledge or mediums concerning the disease and its symptoms, diabetes often goes undetected which has resulted in 1/3 rd of the constituent population being unaware that they possess the disease. The proposed system aims to implement an application to read and analyze medical reports which will generate data that predicts the probabilities of the contraction and onset of diabetes, with insurance of maximum system efficiency and data credibility. Machine learning classification algorithms and optimization techniques have been used to predict diabetes status with maximum accuracy. To extract data from medical reports Optical Character Recognition, Image Processing, and Natural Language Processing have been used