Comparative Study of Machine Learning Approaches in Diabetes Prediction

Diabetes is a common illness that scares people around the world about their health. Biomedical research effort helps in preventing diabetics and treat it in an efficient way. There are lot of traditional systems, but it cannot handle large amount of data and it leads to problems with high levels of complexity and often it was very tedious. This research helps to design a model that can predict the risk of diabetes in patients with acceptable accuracy. Therefore, to identify diabetes in initial stage, this experiment uses machine learning algorithms, namely Random Forest, J48 as well as Multilayer Perceptron. Experiments are carried out on data collected from the UCI machine learning repository that has been gathered from patients. The impacts of all three algorithms are calculated on many scales, such as Accuracy, Precision, Recall and F-Measure. Accuracy is calculated against instances predicted correctly and incorrectly. The results obtained indicates Random forest performs well with the highest precision of 97.5 percent compared to other algorithms but J48 algorithm took minimum time to build the model.