Non- invasive system for heart disease prediction

Abstract

Heart disease (HD) is considered as a potentially fatal disease. Because of multiple contributory risk factors such as abnormal pulse rate, diabetes, high blood pressure, excessive cholesterol, and so on, identifying the condition is difficult. Accurate and timely diagnosis is critical for HD therapy and prevention. As HD is an important reason of deaths in developing nations, the intentions of detecting the risk factors and early sign detection of this disease is important for which the present study has been undertaken. In spite of the various endeavors of conventional works, there is a scope for improvement with regard to accuracy. To attain high prediction ac curacy, as per defined objective 1, in phase-I, the proposed work uses genetic algorithm (GA), particle swarm optimisation (PSO), african buffalo optimisation (ABO) and genetic sine algo rithm (GSA) on the benchmark UCI dataset to remove the unnecessary and redundant features and prevent the features from becoming trapped in local minima. For the classification of spe cific features, the system employs deep learning (DL) based recurrent neural network (RNN), long short term memory (LSTM), and deep progressive attention (DPA-RNN+LSTM) to boost the model’s classification rate. In phase-II, the study proposes penguin optimisation algorithm (POA) for feature optimisation and stacked sparse convolutional based auto-encoder for the classification of HD. As per defined objective 2, further in phase-III, the Raspberry-Pi using the arduino uno and AD8232 sensor framework is used. This made it possible to receive the patient’s body parameters in real time. Since IoT-based data is compiled from a variety of sources, it may be chaotic and noisy. IoT-data mining is employed to help with tasks like defin ing typical links between data components and using them to address prognostication concerns . Also, validate the data for binary and multiclass. To achieve objective 3, in phase –IV, the vascular age, cardiac index, and cardiac risk score is estimated to predict regression values. Fi nally, the classification of heart disease and regression is deployed in the form of flask web page model. Overall, the proposed model is assessed with regard to significant metrics for ensuring its effectiveness than conventional algorithms in HD prognosis.