Leveraging CNN and CNN-LSTM Networks for Automated Diabetes Detection Using Heart Rate Signals

Simple biological signals are unpredictable and hard to plan for. The function of different bodily systems may be revealed via biomedical signals. This makes it difficult to observe these signals and extract temporal data. Use of signal processing methods has allowed for the extraction of valuable information from these signals. A large variety of illnesses may be found and named. Innovative technology that is safe and useful Doctors may find it helpful in identifying patients and therapy due to its exceptional accuracy in spotting abnormalities. control several illnesses. Until recently, machine learning was the only method available for bioanalysis of process data. Methods that rely on machine learning are now in use. Here, we use a deep learning network grouping of CNNs and LSTMs to automate diabetes diagnosis. An analysis of variability in heart rate (HRV) may be used to find diabetes. Up until now, no one has attempted to differentiate between normal HRV and diabetes using deep learning approaches. Its basis as LSTM is a modified continuous neural network (RNN). LSTM customs recollection chunks to address the issue of the disappearing and growing gradient. Once equated to RNNs, LSTMs performs much better when dealing with persistent dependencies. In other words, long short-term memories (LSTMs) may be able to connect seemingly distant events with the here and now. The current record for automated diabetes diagnosis employing HRV is 95.1%, accomplished by means of a CNN-LSTM 3 system with five-fold cross-validation. Regular classifiers and active feature extraction are unnecessary.