Machine-learning models identified key factors linked to TB incidence

{
"What the study found": "The study found that CatBoost, random forest, and gradient boosting were the best-performing models for forecasting monthly tuberculosis incidence in Taiwan, China. The authors also identified population size, sulfur dioxide levels, physician count, normalized difference vegetation index, wind velocity, and precipitation level as the main influences on incidence.",
"Why the authors say this matters": "The authors conclude that the framework and findings provide data support and a decision-making basis for tuberculosis mitigation initiatives on a global scale. They also suggest the results help clarify how environmental and socioeconomic factors relate to tuberculosis incidence.",
"What the researchers tested": "The researchers analyzed data from 19 cities and counties in Taiwan, China from 2014 to 2022. They used four machine learning models and four deep learning models to forecast monthly tuberculosis incidence using 12 drivers, then applied post-hoc explainable machine learning techniques, stepwise regression, and statistical assessments.",
"What worked and what didn't": "CatBoost, random forest, and gradient boosting emerged as the top-performing models. The study also reported nonlinear interactions and threshold effects between the identified determinants and tuberculosis incidence, and stepwise regression was used to find a model configuration that kept high predictive accuracy while reducing the number of drivers.",
"What to keep in mind": "The summary does not describe detailed model performance values or validation results. It also does not provide limitations beyond the study scope of Taiwan, China, 2014 to 2022, and monthly incidence forecasting."
}

Key points

• CatBoost, random forest, and gradient boosting were the top-performing models.
• Population size, sulfur dioxide, physician count, vegetation index, wind velocity, and precipitation were the main influences on tuberculosis incidence.
• The study examined 19 cities and counties in Taiwan, China from 2014 to 2022.
• The authors reported nonlinear interactions and threshold effects between key factors and tuberculosis incidence.
• Stepwise regression was used to reduce the number of drivers while keeping high predictive accuracy.