A State-of-the-Art AF-HIDOP: A Model for Daily Covid-19 Infectious Disease Outbreak Spread Risk Prediction

Abstract

The pandemic produced by the COVID-19 virus has resulted in an estimated 6.4 million deaths worldwide and a rise in unemployment rates, notably in the UK. Healthcare monitoring systems encounter several obstacles when regulating and anticipating epidemics. The study aims to present the AF-HIDOP model, an artificial neural network Fast Fourier Transform hybrid technique, for the early identification and prediction of the risk of Covid-19 spreading within a specific time and region. The model consists of the following five stages: 1) Data collection and preprocessing from reliable sources; 2) Optimal machine learning algorithm selection; 3) Dimensionality reduction utilising principal components analysis (PCA) to optimise the impact of the data volume; 4) Predicting case numbers utilising an artificial neural network model, with 52% accuracy; 5) Enhancing accuracy by incorporating Fast Fourier Transform (FFT) feature extraction and ANN, resulting in 91% accuracy for multi-level spread risk classification. The AF-HIDOP model provides prediction accuracy ranging from moderate to high, addressing issues in healthcare-based datasets and costs of computing, and may have potential uses in monitoring and managing infectious disease epidemics. The procedure and results of phases two and three will be explained briefly in this study; however, identifying the performance of ML algorithms before and after tuning is essential. Hence, a second part will follow this article to elaborate on phases two and three.