AI/ML-Based Diabetes Application using Hybrid Grey Wolf and Dipper Throated Optimization Algorithm

The application of Deep Learning (DL) in diagnosing chronic epidemiological disorders such as diabetes mellitus has become crucial due to the widespread occurrence of this disease worldwide. However, the diagnosis of diabetes faces many challenges such as missing values, high dimensionality of features, and accuracy issues. This paper proposes a Grey Wolf and Dipper Throated Optimization (GWDTO) algorithm for feature selection to address these challenges. The GWDTO algorithm selects relevant features and reduces dimensionality. Initially, data is obtained from the Pima Indian Diabetes Database (PIDD), wherein pre-processing involves handling missing values and normalizing the data by scaling it between the range of 0 and 1. Feature selection using the GWDTO algorithm balances exploration and exploitation to better explore the feature space. The Convolutional Auto encoder (Conv-AE) is then used for classification, after which encoding and decoding of the input data is carried out to capture the underlying structure of the data. The GWDTO algorithm ultimately achieves high accuracy in diabetes diagnosis when compared to the existing techniques namely, Convolutional Neural Networks (CNN) and Naïve Bayes. The proposed method achieves a high accuracy of 99.10% on the PIDD dataset.