الفهرس 
AbstractOnly 14 pages are availabe for public view
 |  | from | 77 |  |  |
| | | from | 77 | | |
Abstract
The goal of this research was to develop a genetic risk prediction model. Diabetes, as previously said, affects a sizable portion of the human population. If left unchecked, it will pose a significant threat to the entire world. The proposed study, On the PIMA dataset, we tested numerous classifiers and demonstrated that data mining and machine learning techniques may reduce risks while enhancing efficiency and accuracy. The PIMA Indian dataset result outperforms other solutions provided on the identical data that use machine learning approaches.
The classifications (DT, ANN, NB, and DL) achieve an accuracy of 91-98%range, which is significantly greater than current methodologies. With a 98% accuracy rate, DL is regarded as the most powerful and appealing of the four potential classifiers for illness treatment.
Diabetic retinopathy can be one of the most hazardous symptoms of diabetes mellitus; failing to diagnose and treat it on time can lead to significant vision loss or even blindness. Diabetic retinopathy, on the other hand, may be prevented with frequent screening and treatment, avoiding irreparable blindness. As machine learning and artificial intelligence technologies progress, many machine learning approaches are being applied in the healthcare sector to aid physicians with regular therapy and diagnosis.
As a result, the ECA is provided in this study for feature maps. In comparison, the ECA-Net DCNN model combines the ECA and Efficient-Net I. This study proposes a convolutional kernel size adjustment methodology for attempting to extract network channel correlation during the disease feature extraction stage, which also allows ECA-Net to creatively change the convolutional kernel size in various tasks, going to allow the system to carefully consider the likeness between feature map channels and improve the outcomes. They were suggested for the quick detection of diabetic retinopathy.
5.2 Future Work
In the future, we intend to create a trustworthy network, like an application or a webpage, that can use the proposed DL technique to assist dataset providers in diagnosing diabetes early.
In the instance of diabetic retinopathy, we want to integrate the ECA learning algorithm with additional deep learning techniques in the future to increase the model’s ability to detect minor variations across classes, allowing ECA-Net to be employed in a broader range of circumstances.