Enhanced Multi-Stage Guided-Attention Mechanism for Traffic Accident-Based Patient Safety Prediction Model

Abstract

Abstract The patient safety prediction model is required for analysis by considering the traffic accident database. In this paper, develop an Enhanced Multi Stage Guided Attention Mechanism (EMSGAM) for a traffic accident-based patient safety prediction model. Patient safety is analyzed by traffic accident data. This data is containing various parameters of age, gender, BMI, hypertension, diabetes, deficiency, depression, and so on. The Min-Max normalisation method is first used to normalise the dataset. Then, with two classes of patients—low-risk and high-risk—the gathered database is used to predict patient safety. The pre-processed data is then forwarded to the feature selection stage, where the necessary features are picked out of the input features. Correlation-based feature selection (CFS) is used to pick the features. The Multi Stage Guided Attention Mechanism (MSGAM) receives the data in order to classify it for patient safety. Long Short-Term Memory (LSTM), the self-attention mechanism, and the Variational Autoencoder (VAE) model are all combined to create the proposed classifier. In this architecture, the optimal weighting parameter is chosen by using Improved Tasmanian Devil Optimization (ITDO). The optimization process is enhancing the performance of the classifier. Based on this evaluation, patient safety is analyzed by considering traffic accident data. This dataset is considered for analysing low-risk and high-risk patients. Performance is assessed using performance measurements, and the suggested is implemented in MATLAB.