Predicting Future Pandemic Waves Using LSTM and Population Mobility Data

Predicting future waves of infectious diseases is crucial for effective public health planning and timely intervention. This study proposes a data‑driven approach for forecasting regional outbreaks by integrating epidemiological data, population mobility patterns, and environmental variables using a Long Short‑Term Memory (LSTM) deep learning model. Daily COVID‑۱۹ case data from the Johns Hopkins University dataset are combined with mobility indicators such as workplace, retail, and residential movement, as well as historical weather variables including temperature and humidity. After preprocessing and feature engineering, the integrated time‑series dataset is used to train an LSTM model designed to capture temporal dependencies in disease transmission dynamics. The model takes historical data from previous days as input and predicts the number of future cases over a short forecasting horizon. Experimental evaluation using metrics such as Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) demonstrates that incorporating mobility and environmental features improves prediction accuracy compared to models relying solely on epidemiological data. The results indicate that human mobility patterns can enhance early detection of upcoming infection waves. This approach can support public health authorities by providing early warning signals, enabling better resource allocation, and assisting policymakers in implementing timely preventive measures.

کلیدواژه ها:

Pandemic Forecasting ، LSTM Neural Networks ، COVID‑۱۹ Dataset ، Human Mobility Data ، Time‑Series Prediction ، Epidemiological Modeling ، Deep Learning ، Public Health Surveillance