Prediction of Diabetes with its Symptoms Based on Machine Learning

Abstract

As the destruction of diabetes is significant to the whole world, we want to focus on it and extract useful information from the correlation between symptoms and disease. The dataset obtained from UCI is the fundamental resource for the research. In order to ensure the accuracy of the project conclusions, three different approaches were used to verify each other: literature analysis, data analysis and machine learning. Literature part mainly contains previous work and large quantities of medical research done on diabetes. Data analysis included data preprocessing and visualization so as to unfold the concealed information of the dataset. Machine learning is to use the inspiration from the previous two parts to attain a suitable model for diabetes prediction. The project finally provides knowledge of different symptoms of diabetes and their relation with diabetes. It also elaborates how symptoms can be used to predict disease. Finally, we put forward suggestions for the prevention of diabetes and monitoring of potential disease.