Abstract
Diabetes mellitus is one of the most widespread chronic diseases affecting millions
of people worldwide. Early prediction and diagnosis can help in controlling
complications and improving quality of life. In this paper, we propose a machine
learning–based approach for diabetes prediction using patient health records.
Various classification algorithms such as Logistic Regression, Decision Tree,
Random Forest, Support Vector Machine (SVM), and Naïve Bayes were applied on the
PIMA Indian Diabetes dataset. The performance of the models was evaluated using
accuracy, precision, recall, and F1-score. Experimental results show that ensemble
models such as Random Forest achieved the highest accuracy, making them a reliable
choice for medical decision support systems.
Keywords — Diabetes Prediction, Machine Learning, Classification, Random Forest,
Healthcare
1. Introduction
Diabetes is a chronic disease that occurs when the body cannot properly regulate
blood sugar (glucose) levels. The World Health Organization (WHO) reports that the
prevalence of diabetes is steadily increasing, posing a major global health threat.
Early detection is crucial for preventing severe complications such as kidney
failure, blindness, and cardiovascular diseases.
Traditional diagnostic methods are often time-consuming and may require invasive
procedures. With the advent of machine learning, it is possible to develop
intelligent systems that can predict the likelihood of diabetes based on clinical
and lifestyle features. This paper focuses on building and evaluating predictive
models for diabetes using machine learning algorithms.
2. Literature Review
Several studies have applied machine learning techniques for predicting diabetes:
Smith et al. (2019) applied logistic regression and achieved moderate accuracy.
Khan & Gupta (2020) demonstrated that Random Forest classifiers outperform
traditional models.
Deep learning models have also been explored, but require larger datasets for
effective training.
From the literature, it is evident that ensemble learning methods show better
predictive performance compared to individual classifiers.
3. Methodology
3.1 Dataset
The PIMA Indian Diabetes Dataset from the UCI Machine Learning Repository was used.
It consists of 768 records with 8 medical predictor variables such as:
Number of Pregnancies
Glucose Level
Blood Pressure
Skin Thickness
Insulin
BMI (Body Mass Index)
Diabetes Pedigree Function
Age
3.2 Preprocessing
Handling missing values
Normalization of data
Splitting dataset into training (80%) and testing (20%)
3.3 Algorithms Used
Logistic Regression
Decision Tree
Random Forest
Support Vector Machine (SVM)
Naïve Bayes
3.4 Evaluation Metrics
Accuracy
Precision
Recall
F1-score
Confusion Matrix
4. Experimental Results
Algorithm Accuracy Precision Recall F1-Score
Logistic Regression 77% 75% 72% 73%
Decision Tree 74% 71% 70% 70%
Random Forest 82% 80% 79% 79%
SVM 79% 77% 75% 76%
Naïve Bayes 76% 74% 73% 73%
Random Forest performed the best with an accuracy of 82%.
5. Discussion
The results indicate that ensemble models such as Random Forest provide better
predictive accuracy compared to single classifiers. Logistic Regression and SVM
also showed promising results. Future work can focus on integrating deep learning
models, feature selection techniques, and larger datasets to further improve
prediction accuracy.
6. Conclusion
This study demonstrates that machine learning algorithms can effectively predict
diabetes based on clinical features. Among the models tested, Random Forest
achieved the highest performance. Such predictive systems can assist healthcare
professionals in early detection, leading to timely medical interventions.