DT-DL Based Hybrid Approach for Early Detection of Diabetes Using PIMA Dataset

According to the International Diabetes Foundation (IDF) 2019 report, 463 million people are living with diabetes, which is likely to increase to 700 million by 2045. Diabetes Mellitus is a metabolic disease with a persistent increase in pervasiveness. Therefore, it is one of the most severe challenges in developed and developing countries. Detection of diabetes is critical in the early phase so that the progression of the disease can be stopped at a definite complication. Therefore, it is mandatory to build an automatic tool that can predict diabetes at an early stage. The machine learning approach has been proven promising for detecting diabetes accurately and is hence used in many diabetes detection processes. This study presents the practical hybrid approach by combining two machine learning algorithms (Decision tree, Deep learning) that may provide prediction and early detection of diabetes securely and effectively. This study's result confirms that the proposed framework can be applied for early diabetes detection with ID3 and J48 algorithms to analyze the performance. The accuracy achieved by these algorithms was recorded as 94.6% and 88.2%, respectively. Additionally, results are compared with the proposed DT-DL hybrid classification approach, which shows that our proposed approach outperforms other traditional approaches with an accuracy of 96.62%. The outcome of this study shows that the hybrid approach of DT-DL provides the most promising results with the best-extracted features.