DOI: https://doi.org/10.59321/BAUETJ.V4I1.9
AUTHOR(S)
A. K. Soykot Azad Snigdho1, Nazmul Hussain1*, Md. Abdul Hamid2, Nasirul Mumenin1, Tamanna Jannat1, Anika Tahsin1, Md. Rajib Ali1, Partha Pratim Debnath1
ABSTRACT
Diabetes mellitus is a cluster of conditions that impact the body’s utilization of glucose, a vital energy source for the cells of muscles and tissues. International Diabetes Federations (IDF) shows that almost 382 million people are living with diabetes. The goal of this research is to predict diabetes by supervised machine learning algorithms. The result is then compared with the deep learning approaches. The conventional machine learning algorithms are used here i.e., logistic regression (LR), gradient boost (GB), decision tree (DT) and random forest (RF). Then deep learning (DL) method implement to predict and detect diabetes through neural network. The research is done with the Pima Indians Diabetes dataset which is publicly available. This dataset consists of eight input parameters with 768 samples where 268 samples for diabetic and rest of them are non-diabetic patients. The accuracy was obtained using the LR, GB, DT, RF and DL are 86%, 93%, 91.2%, 95% and 94.2% respectively. The accuracy shows that random forest gained better performance than the logistic regression, gradient boost, decision tree and deep learning.