Intelligent Framework for Classification of Influenza Severity for Respiratory Disease Management

Abstract

Introduction: Influenza is a highly contagious respiratory illness that poses significant public health risks, with severity varying across individuals. Accurately assessing the severity of influenza is crucial for timely intervention and resource allocation. This study explores the application of machine learning, specifically Random Forest regression, to predict influenza severity using key physiological indicators. Problem Statement: Traditional clinical assessment methods for influenza severity rely heavily on subjective evaluations and may not fully capture complex interactions among multiple health indicators. There is a need for a robust, data-driven approach that can objectively predict severity and inform early clinical decisions. Aim: The primary objective of this research is to develop and evaluate a machine learning model capable of predicting the severity index of influenza based on key physiological features, thereby enabling early detection and improved patient management. Methodology: A comprehensive datasetsourced from Kaggle was used, including featuressuch as body temperature, respiratory rate, oxygen saturation, cough intensity, fatigue level, and white blood cell (WBC) count. After data preprocessing, the dataset was split into training (80%) and testing (20%) sets. The Random Forest regression model was selected for its capacity to handle nonlinear feature interactions and prevent overfitting. Model performance was assessed using Mean Squared Error (MSE) and the coefficient of determination (R²). Feature importance analysis and visualizations, including regression plots and scatter visualizations, were conducted to interpret the model outputs. Results: The Random Forest regression model achieved an MSE of 0.0173 and an R² value of 0.6630, indicating that the model explained approximately 66.3% of the variance in severity index. Key features such as respiratory rate, oxygen saturation, body temperature, and WBC count demonstrated high predictive significance. Higher respiratory rates and lower oxygen saturation were strongly associated with increased severity, while elevated body temperature and WBC count contributed to symptom progression. Visualizations reinforced these relationships, providing a clear view of the feature-severity dynamics. Recommendation: The predictive capabilities of the developed model suggest its potential utility as a clinical decision-support tool for early influenza severity assessment. It is recommended that healthcare facilities consider integrating such models into patient monitoring systems for real-time severity evaluation. Future research should explore larger, more diverse datasets and investigate ensemble learning techniques to enhance predictive accuracy. Additionally, the integration of longitudinal patient data could further refine severity trajectory predictions, offering a more dynamic approach to disease management. This study underscores the transformative potential of machine learning in respiratory disease management, offering a pathway toward more proactive, personalized, and data-driven healthcare solutions.