Enhancing Automated Diabetes Diagnosis by the Use of More Advanced Spectral Characteristics in Heart Rate Readings

Biomedical signals may reveal body component function. Basic biological signals are unpredictable and irregular. This makes temporal data from such signals is difficult to get via observation. Signal processing technologies have been used to extract meaningful information from these signals. so many illnesses may be found and identified. Technology that is non-invasive and beneficial Its high anomaly detection accuracy may help physicians identify and treat patients. control many illnesses. Machine learning was formerly the standard for bio analysis of signals from processes. Currently, deep learning approaches are applied. ECG signals show how the ANS maintains a regular heartbeat. ECG signals are non-linear and unrelated to nature. Some disorders may be more effectively diagnosed using ECG analysis. Diabetes, a serious condition, may be detected by an ECG. Diabetes is incurable. Diabetes may lead to cardiovascular disease, renal disease, and stroke if not effectively managed with medicine and insulin. Early diagnosis improves diabetic management. This thesis aims to illustrate that electrocardiogram (ECG) heart rate variability (HRV) values can detect diabetes. More sophisticated spectral (HOS) properties are required for the interpretation of non-linear HRV data. This research combines regular and diabetic ECG data to uncover HRV-generated HOS characteristics. These collected attributes were given to seven classification algorithms, including naive bayes, SVM, and others. Classifier performance was measured by accuracy, positive prediction value, sensitivity, and specificity. The results are compared to other methods and two parameters of the at its peak, the classifier using SVM achieves a predictive value (PPV) of 98.3% and a specificity rating of 98.4%, with an accuracy of 86.5%.