Towards Social Group Optimization and Machine Learning Based Diabetes Prediction

Diabetes has become a major health concern among people around the world. It may reduce the life span, and enhances the probability of various kind of cardiovascular diseases. Prediction of diabetes may provide an alarm to the people who are required to check their health status. It is a very challenging task as medical data is very much saturated and complex. In this work, we have predicted diabetes from several lifestyle parameters such as BMI, age, pregnancy and symptoms such as weight loss, visual blurring, weight loss, and so on. Several machine learning algorithms were used to predict diabetes. These machine learning algorithms were further optimized using Particle Swarm Optimizer and Social Group Optimizer. Social Group Optimized Gradient Boosted Classifier (GBC) achieved an accuracy of 71.85% in predicting diabetes from lifestyle parameters. The proposed architecture achieved 98.26% accuracy in case of prediction from symptoms using Social Group Optimized Random Forest Algorithm.