Scribd
Upload
21 views17 pages

Comparative Analysis of Diabetes Prediction Using Machine Learning

This document compares different machine learning approaches like SVM, MLP, and CNN for detecting diabetes mellitus. It discusses types of diabetes, methods for diagnosis, and compares the performance of different algorithms on diabetes prediction datasets. SVM performed best on one dataset while DNN worked best on another. Increasing dataset size is suggested for more complex predictions.

Uploaded by

Ahmed ragab Galal
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
21 views17 pages

Comparative Analysis of Diabetes Prediction Using Machine Learning

This document compares different machine learning approaches like SVM, MLP, and CNN for detecting diabetes mellitus. It discusses types of diabetes, methods for diagnosis, and compares the performance of different algorithms on diabetes prediction datasets. SVM performed best on one dataset while DNN worked best on another. Increasing dataset size is suggested for more complex predictions.

Uploaded by

Ahmed ragab Galal
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 17

Comparative Analysis of Diabetes

Prediction Using Machine


Learning
By / Yossef Haithem _ Omar Ahmed Shinawy _ Ahmed Ragab Galal
Amr mousa Nazir _ Mohamed Abdalla Boghdady
Paper's goal | General information

compare between different As reported in 2017, there were


approaches (SVM, MLP, CNN) for 245 million people affected with
detection of diabetes mellitus diabetes. Expected to be 625
using , this computer-based million in 2045.
detection helps doctors diagnose
the disease in the early stage.
what is diabetes?

▪ Pancreas: Produces the hormone insulin.


▪ Insulin’s Role: Allows nutrients from the food we
eat (including sugars, carbohydrates, and fats) to
enter cells in the body.
▪ Energy Production: Cells use these nutrients to
produce energy.
▪ Importance of Insulin: Without insulin, glucose
levels rise, leading to health issues.
Blood sugar level
Types of Diabetes

Type 1 Diabetes Type 2 Diabetes

- occurs in child and adolescents - mostly occurs in adults


- it occurs when there is little - occurs when pancreas does
amount of insulin or no insulin not produce insulin
-should take insulin injections or - can be controlled by doing
tablets as daily routine physical activities and taking
healthy diet or some oral drugs
Gestational Diabetes(GDM)

▪ can be seen in pregnancy women


▪ GDM is a condition in which placenta releases a
hormone, and it prevents the body from using
insulin effectively.
▪ can cause complications for both mother and child
▪ GDM will disappear after pregnancy, and there are
high chances that mother and child may get type 2
diabetes in future
Methods to diagnose diabetes

▪ Fasting Plasma Glucose(FPG)


▪ Oral Glucose Tolerance Test (OGTT)
▪ A1C
Fasting Plasma Glucose(FPG)

▪ should be held before breakfast and make sure the person


should not consume any food except water before 8 h

▪ The plasma glucose value 126 mg/dL or higher leads to


diabetes

▪ Prediabetes values are 100–125 mg/dL and below to this


level has no diabetes.
Oral Glucose Tolerance Test (OGTT)

▪ blood sugar levels before and after a


special sweet drink within a span of two
hours
▪ declares how diabetic body processes
sugar
▪ The oral glucose tolerance test value 200
mg/dL or higher leads to diabetes
▪ Prediabetes values are 140–199 mg/dL
and below to this level has no diabetes
A1C

- The average blood sugar levels for past two to


three months are measured
- A1C test value 6.5% or higher leads to
diabetes
- Prediabetes values are 5.7–6.4 % and below
to this level has no diabetes
- The main advantage of this test is no need to
fast
The comparison

1. Supervised vs. Unsupervised Learning: 3. Early Diabetes Prediction with Deep Neural
1. Researchers have explored both supervised and Networks:
unsupervised machine learning methods. 1. Early detection of diabetes is crucial for
2. Unsupervised methods provide quicker effective treatment.
outputs, while supervised methods offer higher 2. Researchers proposed a model that predicts
performance. diabetes and determines its type.
3. The choice depends on the specific task and 3. Deep neural networks were employed, using
requirements. large datasets (over 1000 records).
4. Achieved 94.02% accuracy with a diabetes
dataset and 99.41% training accuracy with the
Pima Indians diabetes dataset.
1. Hybrid Approach (Fused Machine Learning):
1. Some studies utilize a hybrid approach 4. Hybrid Technique for Diabetic Prediction:
combining different techniques. 1. This methodology combines three
2. Example: In one approach, an initial stage algorithms: Support Vector Machine
involves SVM and a one-layer neural network. (SVM), Random Forest (RF), and Convolutional
The output from this stage serves as input for Neural Network (CNN).
a fuzzy model, resulting in accurate predictions. 2. Accuracy was optimized by performing multiple
experiments and averaging the results.
Result and Discussion
Tables 1&2
Comparative Analysis of Diabetes
Tables 3&4
Conclusion

• Machine Learning Algorithms: The paper compares various algorithms


for diabetes prediction, focusing on SVM, ANN, DNN, and CNN.

• Performance Metrics: Algorithms were evaluated based


on sensitivity, specificity, and accuracy.

• Datasets Used: Two datasets were used for comparison: the UCI
dataset and the PIMA dataset.

• Results:
• SVM yielded the best results for the UCI dataset.
• DNN was most effective for the PIMA dataset.

• Future Work: Suggests increasing the dataset size for higher accuracy in
complex predictions.
Thanks

You might also like