Feature Selection with PSO and Convolutional Neural Network with Long Short-Term Memory for Medical Application

Abstract

Abstract The presence of cardiovascular disease is the main cause of fatalities and illnesses throughout. Predicting heart disease prognosis in medical data analytics is quite difficult. The spectacular amount of unstructured data produced by search has demonstrated that significant features are crucial in enhancing the effectiveness of machine learning models. In this study, the dataset of hospitalised patients' heart failure survivors is analysed. It is important to discover critical characteristics and effective deep-learning methodologies to increase the accuracy of survival prediction for cardiovascular patients. This research is approached by using K-Means segmentation for partitioning data. After segmentation, the highest-ranking features chosen by Particle Swarm Optimization (PSO) are used to train machine learning models. Finally, trained data of patients are classified using a hybrid technique combining Convolutional Neural Network (CNN) and Long-Short Term Memory (LSTM). The classification attains an overall accuracy of 94.63%, thereby supporting experts to predict disease earlier. The examination of the proposed is accomplished through Python Software and the outcomes reveal that the hybrid classification technique performs better metrics when contrasted with state-of-art approaches.