Artificial Intelligence-based Learning Techniques for Diabetes Prediction: Challenges and Systematic Review

Abstract

Diabetes is one of the most deadly and chronic diseases which cause an increase in blood sugar. If diabetes remains untreated and unidentified many difficulties may arise due to that. The tedious work is in identifying the process which results in visiting the clinic and consulting the doctor. But this tedious work has been solved with the rise in the approaches used by machine learning. Over recent years a plenty of growth has been seen doing over health informatics by focusing on the technology of the presentation, generation and application of clinical information in health care. With the motive of improving outcomes of health for patients and creating efficiency in health professions a Healthcare Informatics or eHealth solutions, has empowered the accessibility of clinical data through computer networks or cloud computing. Its accessibility and understanding have become easier with language technologies. This paper gives a comprehensive prospect of work accomplished to develop a model that can predict the possibility of diabetes in patients with extreme accuracy. Therefore, various machine learning classification algorithms are used for detecting diabetes. In this paper, the author has studied various machine learning classification algorithms, namely genetic algorithm, decision tree, random forest, Logistic regression, SVM and Naive Bayes. Experiments are carried out on Pima Indians Diabetes Database (PIDD) which is track down from the UCI machine learning repository. Further author has done the comparison among various performances of all the different algorithms. The performances are categories of various measures like Precision, Accuracy, F-Measure and Recall. The paper helps in identifying the algorithm to classify the risk of diabetes. Different techniques were applied to the algorithms for improving the robustness. Additionally, the findings suggest that the best performance of disease risk classification is done with the help of a genetic algorithm.