An Integrated Approach for Diabetes Detection Using Fisher Score Feature Selection and Capsule Network

Abstract

This paper investigates how the Fisher score feature selection approach can be used with capsule networks for diabetes detection. It also evaluates how well this algorithm works based on a number of evaluation parameters. The selected features using Fisher score method was then employed to train a capsule network model. Accuracy (94%), precision (94%0, recall (94%), F1 score (94%), and other performance evaluation metrics were thoroughly analyzed to determine the algorithm's efficacy. The results demonstrated that the combination of Fisher score feature selection and capsule networks yielded promising performance in diabetes detection. The selected features effectively captured the relevant information necessary for accurate classification The capsule network model was very accurate, which shows that it could be a good tool for diagnosing diabetes. Also, the accuracy and recall values showed that the algorithm could correctly place both positive and negative cases of diabetes, minimizing the risk of misdiagnosis. By merging the Fisher score feature selection approach with capsule networks, this research study contributes to advancing diabetes detection.