Application research on surveillance and predictive modeling of respiratory diseases in Baise City based on...

Respiratory diseases are increasingly influenced by meteorological variability, yet few studies have applied high-resolution environmental data and advanced forecasting models to quantify this relationship in climate-sensitive regions such as southern China. This study aims to characterize how climate variability drives respiratory disease dynamics by quantifying the influence of meteorological factors and constructing predictive models for early intervention. It further seeks to advance climate-resilient health surveillance through the application of machine learning and time-series forecasting techniques. A longitudinal dataset comprising daily meteorological indicators (temperature, humidity, precipitation, wind speed, atmospheric pressure, air quality index) and annual respiratory disease incidence was analyzed. Feature engineering, dimensionality reduction via Principal Component Analysis (PCA), and temporal lags were applied. Predictive models—Random Forest, Support Vector Regression (SVR), Long Short-Term Memory (LSTM), and Seasonal AutoRegressive Integrated Moving Average with eXogenous regressors (SARIMAX)—were evaluated using coefficient of determination (R²), Mean Absolute Error (MAE), and Root Mean Square Error (RMSE). Variable importance and Pearson correlation were used to interpret model behavior and epidemiological relevance. Respiratory incidence showed marked interannual fluctuations, with notable peaks in 2014, 2018, 2022, and 2024. Random Forest identified total precipitation (27%), average temperature (22%), and relative humidity (18%) as the most influential predictors. Air quality and wind speed also contributed to model accuracy. LSTM and SARIMAX models achieved high fidelity in forecasting both long-term and weekly trends, with low residual variance and strong temporal alignment with observed case data. This study provides strong empirical evidence that respiratory disease patterns in Baise City are closely linked to meteorological conditions. The integration of machine learning and time-series modeling establishes a scalable framework for real-time forecasting and early detection of climate-sensitive health risks. These findings advance the frontier of climate-health analytics and offer actionable pathways for embedding environmental intelligence into public health policy, resource planning, and resilience-building strategies in vulnerable regions.