Hybrid ARIMA-LSTM Model for Real-Time Influenza Monitoring and Early Prediction
A. Krishnaveni, K. Gowtham · 2025
Influenza is a highly contagious viral infection capable of widespread epidemic dispersion with catastrophic effects on community health. Rapid detection of influenza outbreaks is essential for limiting virus transmission and health system burden. This research proposes a hybrid artificial intelligence (AI) and machine learning (ML) based system aimed at the surveillance of influenza, combining both auto-regressive integrated moving average (ARIMA) for longitudinal periodic trends and long short-term memory (LSTM) networks for unpredictable short-term fluctuations. This problem is handled by one system that breaks down the process using time-series surveillance data and external environmental factors to predict the outcome. Including other environmental features (e.g., temperature, humidity, and Air Quality Index (AQI)) led to a substantial performance improvement for the prediction. The identification of early diseases can happen because Google Search Trends monitors behavioural indicators that warn about flu symptom appearances. The initial stage of data processing applies both feature selection and augmentation along with normalization before executing clean-up procedures. The combination model uses RMSE, F1 Score, Accuracy, Precision, and Recall for performance evaluation. Results show that an accuracy level of 96.3% was obtained using the proposed combination model, which is better than individual models. The results of the research prove the potential of AI-based systems to create a public health warning system as well as the analysis of influenza epidemic threats.