Development and Implementation of a Diabetes Prediction System using Logistic Regression

Creative Commons License

Lawrence M. O., Ibrahim A. M., AWOTUNDE J. B., Abiodun M. K., Alfa P. E., Adebiyi A. A.

NIPES - Journal of Science and Technology Research, cilt.7, sa.1 Special Issue, ss.384-392, 2025 (Scopus)

Özet

Diabetes indeed is a major worldwide health concern, and early detection and intervention expedite disease control and treatment. This work presents a Diabetes Prediction System that classifies individuals into diabetic and non-diabetic categories based on important health indicators. Featuring an ML algorithm that analyzes different health characteristics and risk factors, this system predicts diabetes with accuracy. Blood pressure, glucose levels, age, skin thickness, insulin level, number of pregnancies, and diabetes pedigree function are considered as some of the risk factors. The algorithm is trained on a dataset of 768 patients taken from Kaggle (Pima Indians Diabetes Datasets), representing diabetic and non-diabetic cases. A proper preprocessing step rectifies absent values and normalizes features within the dataset. The results reveal that the LR model type attained 81% accuracy, 74% precision, 68% recall, and 71% F1 score. In particular, logistic regression predicted 35 true positives and 91 true negatives while producing 12 false positives and 16 false negatives. It also means that the prediction system shows that an accurate diabetes prediction can be attained using an ML approach, with Logistic Regression being the best in this case. The application of this technology for early detection will help patients for a better future and assist health authorities in making well-informed decisions for the better control of diabetes worldwide.