A Review of Machine Learning Techniques for Diabetes Prediction in Healthcare

Abstract

Diabetes is one of the threatening illnesses to the whole world, however, it is not fatal. Detection of diabetes in the early stages is critical as it assists with decreasing the deadly impacts of it. There have been a lot of research learns about diabetes detection, many of which depend on the Pima Indian diabetes. It is a dataset examining women in the Pima Indian population that began in 1965, where the beginning rate for diabetes is similarly high. Machine Learning (ML) strategies have been used in anticipating diabetes analysis over the most recent years. Even though various ML algorithms were utilized in tackling this problem, there exist several classifiers that are used once or not even once, it must ensure the performance of those classification algorithms to predict diabetes. The expanding unpredictability of this issue has propelled analysts to explore advanced Deep Learning techniques. The suggestion is to apply those classifiers in the prediction of diabetes and alter them by creating hybrid models. The most noteworthy precision accomplished so far was 95.1% by a hybrid model, which is a combination of CNN and LSTM. However, utilization of these models is computationally expensive, often require a huge memory. The majority of the research studies done before primarily concentrated on increasing the model performance, while the best research done on many traditional approaches is not found. So, in this paper, we have given a comprehensive review of the research on predicting diabetes using ML techniques.