DEPARTMENT OF CSE(AI&ML)

Real Time Research Project On

Diabetes Prediction using Machine Learning

PRESENTED BY: UNDER THE GUIDANCE OF:

• S. Pranaykumar (227r1a66j8) M. Balaji
• B. Vaishnavi (227r1a66d9)
• Y. Yugendhar (227r1a66k3)
(Asst. Professor )
 Table Of Contents
• 3. Abstract
• 4. Existing system
• 5. Proposed system
• 6. Software and hardware requirements
• 7. System Architecture
• 8. Usecase diagram
• 9. Class diagram
• 10. Sequence diagram
• 11. Algorithms
• 12. Dataset description
• 13. Output screenshots
• 14. Performance metrics
• 15. Result analysis
• 16. Conclusion
• 17. Future Scope
• 18. Any Queries
 Abstract
1. Day by day the cases of diabetic diseases are increasing at a rapid rate and it’s very Important
and concerning to predict any such diseases beforehand.

2. This diagnosis is a difficult task i.e. it should be performed precisely and efficiently. The
research paper mainly focuses on which patient is more likely to have a diabetic disease based
on various medical attributes.

3. We prepared a diabetic disease prediction system to predict whether the patient is likely to be
diagnosed with a diabetic disease or not using the medical history of the patient.

4. We used different algorithms of machine learning such as SVM to predict and classify the
patient with diabetic disease.
 Existing System

1. There was an existing system called Diabetic Disease prediction using Machine
Learning. They have used the data set which was very small i.e it contains around 100-
120 entries and 5-6 parameters to check. The existing model used to possess an
accuracy rate of 77%.

2. GUI of existing model was Static . This model will ask for 4-5 kind of symptoms and
try to predict the chances of getting Diabetic disease .
 Proposed System
1. Our System basically provides the enhanced results , data set got improved
which consists around 400 entries of data .

2. This System has an accuracy score of 88% which can be trustworthy ,

NULL, duplicate values are removed.

3. Model is designed in such a way that it is capable of providing the best and
accurate results.

4. In this process we have used algorithms like SVM to achieve high

Accuracy.
Hardware Requirements
Processor Intel i5

Mouse Two or Three Button

Mouse
RAM 8GB.

Hard Disk 500GB

Key Board Standard Windows

Keyboard.

Monitor Any LED type

 Software Requirements

Operating System Windows 7 or Windows 11

Programming Language Python

Libraries Numpy , pandas ,sklearn

Dataset diabetic.csv from google

System Architecture
Use Case Diagram
Class Diagram
Sequence Diagram
 Algorithms:
SVM is a powerful supervised machine learning algorithm used for classification and
regression tasks. In the context of diabetes prediction, SVM can classify whether a patient has
diabetes based on various health metrics (features).

Steps for Using SVM in Diabetes Prediction:

1. Data Collection: Gather a dataset with features such as glucose level, BMI, age, etc., and labels
indicating whether the patient has diabetes.
2. Data Preprocessing: Clean and prepare the data (handle missing values, normalize features,
split into training/testing sets).
3. Model Training: Use the training set to train the SVM model. The SVM algorithm will find the
hyperplane that best separates diabetic from non-diabetic cases.
4. Model Evaluation: Evaluate the model on the test set using metrics like accuracy, precision,
recall, and F1-score to ensure it performs well.
5. Hyperparameter Tuning: Adjust parameters such as the kernel type (linear, polynomial, RBF)
and the regularization parameter (C) to improve performance.
 Dataset Description
Pregnancies: Number of times pregnant
Glucose: Plasma glucose concentration (2 hours in an oral glucose tolerance test)
Blood Pressure: Diastolic blood pressure (mm Hg)
Skin Thickness: Triceps skinfold thickness (mm)
Insulin: 2-Hour serum insulin (mu U/ml)
BMI: Body mass index (weight in kg/(height in m)^2)
Age: Age in years
Target Variable:

Outcome: Class variable (0 or 1), where 0 indicates non-diabetic and 1 indicates diabetic.
Output Screenshots:
 Performance Metrics
1. Accuracy
Accuracy is the ratio of correctly predicted instances to the total instances.
Accuracy=True Positives+True NegativesTotal Instances\text{Accuracy} = \frac{\text{True
Positives} + \text{True Negatives}}{\text{Total
Instances}}Accuracy=Total InstancesTrue Positives+True Negatives​
2. Precision
Precision is the ratio of correctly predicted positive observations to the total predicted
positives. Precision=True PositivesTrue Positives+False Positives\text{Precision} = \frac{\
text{True Positives}}{\text{True Positives} + \text{False
Positives}}Precision=True Positives+False PositivesTrue Positives​
3. Recall (Sensitivity or True Positive Rate)
Recall is the ratio of correctly predicted positive observations to all observations in the actual
class. Recall=True PositivesTrue Positives+False Negatives\text{Recall} = \frac{\text{True
Positives}}{\text{True Positives} + \text{False
Negatives}}Recall=True Positives+False NegativesTrue Positives​
 Result analysis
1. Analyzing the results of a diabetes prediction model involves evaluating several key
metrics to assess its performance. The confusion matrix provides a detailed breakdown of
predictions, distinguishing between true positives (correctly identified diabetic patients),
true negatives (correctly identified non-diabetic patients), false positives (non-diabetic
patients incorrectly identified as diabetic), and false negatives (diabetic patients incorrectly
identified as non-diabetic). These metrics are essential for understanding how well the
model distinguishes between diabetic and non-diabetic individuals.

2. These metrics collectively provide a comprehensive view of the model's performance,

informing further refinements or adjustments necessary for improving its accuracy and
reliability in real-world diabetes prediction scenarios.
 Conclusion
1. Through our project we came to a solution of skipping Diagnosis and letting
people know about their heart diseases instantly.
2. We have achieved an accuracy of 78% and have used machine learning
algorithms like SVM.
3. Neccesity of Physical Diagnosis is reduced for a large extent.
4. Our model follows these specified algorithms and predicts whether the person
is having heart disease or not.
 Future Scope

1. The future of diabetes prediction using machine learning holds immense

potential to revolutionize healthcare by improving early detection,
personalized treatment strategies, and patient outcomes.

2. By embracing these future scopes, researchers and healthcare

practitioners can pave the way for more effective diabetes management
and prevention strategies in the years to come.
Any Queries ?
Thank You!!