Predicting diabetes using
machine learning
algorithms!
Presented by: Supervisor:
Name: Jahidul Islam [Link] Islam
ID: 2018200010041 Lecturer
Department of CSE
Southeast University,
Name: Rajib Das Bangladesh
ID: 2018200010051
Name: Rajesh Khoksi
ID: 2018200010004
Abstract
● Diabetes is a chronic condition affecting a large number of
people worldwide, with significant health consequences.
● Machine learning has the potential to aid in early detection and
personalized treatment of diabetes by analyzing patient data.
● This study aims to evaluate the accuracy of five algorithms for
diabetes prediction and explore opportunities for improvement.
Introduction
● Diabetes: a chronic condition with detrimental health [Link]
estimated 463 million adults worldwide suffer from diabetes.
● Machine learning offers potential solutions for diabetes
management and prediction. By analyzing patient data, it can
aid in early detection and personalized treatment.
● Four algorithms are employed in this study for diabetes
prediction. The accuracy of these models will be evaluated, and
opportunities for improvement will be explored.
Literature Review
Paper title Author Summary Models with
accuracy
Prediction and Detection of Llahaa, O., & Ristab, A. The article examines Naive
Diabetes using Machine Lea (Year) interpretability Bayes(72.5%),
rning.(2021) SVM (75.2%),
approaches for
Decision Tree -
diabetes prediction C4.5
models, highlighting (79.3%),Logistic
the influence of blood Regression(68.9%),
glucose on the Random
model's outcome. Forest(77.8%), LR
(77.8%)
Literature Review
Paper title Author Summary Models
Diabetes Prediction Using KM Jyoti Rani KM Jyoti Rani's research on Decision
Machine Learning "Diabetes Prediction Using Machine Tree(98%),
Learning" aimed to develop a Random
decision-making framework Forest(97.8%),
SVM (77.6%), KNN
(78%)
Diabetes Prediction Using Hakim El Mass The authors aim to explore and Random
Machine Learning Algorith ari, Zineb Sab evaluate machine learning forest(87.66%),
ouri, Sajida M techniques for the analysis and XgBoost(74.01%).
ms and Ontology hammedi and
Noreddine Gh prediction of diabetes.
erabi
Methodology
● Data Collection
● Data Preprocessing
● Splitting the Dataset
● Model Training
● Model Evaluation
Dataset
Dataset
Total data 768
Missing cell 0
Duplicate row 0
Data Processing
● Abnormal Zero value handling using quantiles.
● Data imputation to address missing values.
● Quantile-based replacement for meaningful representation.
● Splitting the data set into 80% into training data set and the 20% for testing.
Logistic Regression
KNN
Random Forest
Naive Bayes
Voting Classifier
Conclusion
1. Random Forest: The random forest model performs exceptionally well with high accuracy,
precision, recall, and F1 score. It is highly recommended for classification tasks.
2. Logistic Regression: Logistic regression demonstrates good performance and is a reliable
choice, especially when interpretability is important due to its coefficient interpretation.
3. Voting Classifier: The voting classifier combines multiple models and provides a reliable
option by leveraging the strengths of different models for improved performance.
4. Naive Bayes: The Naive Bayes model achieves moderate performance and can be
considered as an option for simpler datasets, but may not be the most reliable or
recommended choice for all scenarios.
5. K-Nearest Neighbors (KNN): The KNN model has lower performance compared to other
models and may not be the most reliable or recommended option, particularly for high-
dimensional or imbalanced datasets.
References
1. Llahaa, O., & Ristab, A. (Year). Prediction and Detection of Diabetes using Machine Learning.(
2021)
.
2. "KM Jyoti Rani. Diabetes Prediction Using Machine Learning"
3. Hakim El Massari, Zineb Sabouri, Sajida Mhammedi and Noreddine Gherabi- Diabetes Predicti
n Using Machine Learning Algorithms and Ontology
4. G. PRASHANTH [Reg No: RA1511008010408] DEEPTHI [Reg No: RA1511008010580] UMA SIVAKUMAR [Re
No: RA1511008010505] V. PAVAN KAILASH SREEVATSAN [Reg No: RA1511008010057]. PREDICTION OF D
ABETES USING MACHINE LEARNING
5. Pima Indians Diabetes Database. [Link]
Thank you.