Deep Neural Network Classifier for Analysis of the Debrecen Diabetic Retinopathy Dataset
Diabetic retinopathy (DR) is a serious complication that can occur in individuals who have diabetes. This disease affects the blood vessels in the retina, a part of the eye that is important for vision. Early detection of DR is key to preventing further complications and saving the patient’s vision....
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Indonesian Institute of Sciences
2024-12-01
|
| Series: | Jurnal Elektronika dan Telekomunikasi |
| Subjects: | |
| Online Access: | https://www.jurnalet.com/jet/article/view/640 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Diabetic retinopathy (DR) is a serious complication that can occur in individuals who have diabetes. This disease affects the blood vessels in the retina, a part of the eye that is important for vision. Early detection of DR is key to preventing further complications and saving the patient’s vision. The goal of Diabetic Retinopathy Debrecen Data Set Analysis is to get the best, most accurate results for medical professionals to receive appropriate Diabetic Retinopathy Debrecen prediction results through the stages of data collection, evaluation, and classification. Data is collected from existing secondary sources, then assessed using a deep neural network algorithm with various variations. The classification algorithm in this research uses the Python programming language to measure accuracy, F1-Score, precision, recall, and ROC AUC. The test results show that the accuracy of the deep neural network algorithm is 79.94%, the F1 score reaches 79.16%, the precision is 79.58%, the recall is 79.60%, and the AUC is 79.56%. Thus, based on this research, the deep neural network data mining technique with variations of the four hidden layer encoder-decoder, sigmoid activation function, Adam optimizer, learning rate 0.001, and dropout 0.2 is proven to be effective. When compared with other variations such as decoder-encoder, 3-8 hidden layers, learning rate 0.1 and 0.01, the average difference in values between this variation and the others is 0.07% accuracy, 2.03% F1 score, 0.25% precision, 0.80% recall, and 0.90% AUC. Therefore, the deep neural network algorithm with the variation used shows significant dominance compared to other variations.Diabetic retinopathy (DR) is a serious complication that can occur in individuals who have diabetes. This disease affects the blood vessels in the retina, a part of the eye that is important for vision. Early detection of DR is key to preventing further complications and saving the patient’s vision. The goal of Diabetic Retinopathy Debrecen Data Set Analysis is to get the best, most accurate results for medical professionals to receive appropriate Diabetic Retinopathy Debrecen prediction results through the stages of data collection, evaluation, and classification. Data is collected from existing secondary sources, then assessed using a deep neural network algorithm with various variations. The classification algorithm in this research uses the Python programming language to measure accuracy, F1-Score, precision, recall, and ROC AUC. The test results show that the accuracy of the deep neural network algorithm is 79.94%, the F1 score reaches 79.16%, the precision is 79.58%, the recall is 79.60%, and the AUC is 79.56%. Thus, based on this research, the deep neural network data mining technique with variations of the four hidden layer encoder-decoder, sigmoid activation function, Adam optimizer, learning rate 0.001, and dropout 0.2 is proven to be effective. When compared with other variations such as decoder-encoder, 3-8 hidden layers, learning rate 0.1 and 0.01, the average difference in values between this variation and the others is 0.07% accuracy, 2.03% F1 score, 0.25% precision, 0.80% recall, and 0.90% AUC. Therefore, the deep neural network algorithm with the variation used shows significant dominance compared to other variations. |
|---|---|
| ISSN: | 1411-8289 2527-9955 |