Coughing and Seasonality: A Study Based on Machine Learning and Internet Data Analysis

Cough is a defensive physiological response, serving as a protective reflex of the respiratory system against irritants. The incidence of cough varies across different seasons. Although it's a general consensus, there is currently insufficient data to fully support this conclusion. This study aims to examine the relationship between cough and seasonal transitions using internet-based data based on machine learning. Meteorological data, air pollution data, and search index data for coughing were collected from Internet, with research area in Jinan, China from January 2014 to December 2021. Spearman's rank correlation coefficient was utilized to explore the associations between the aforementioned variables. A distributed lag non- linear model was employed to assess the impact of air pollution on the search index data for cough. Time series annual and monthly decomposition models were utilized to analyze the association between the search index for coughing and seasonal variations. The risk factors of cough were analyzed by machine learning model. The results demonstrated a connection between the search index for coughing and seasonal changes, with coughing reaching its peak during cold seasons. Specifically, the search index for cough showed an upward trend from July to January of the following year, while it experienced a downward trend from January to June. The monthly decomposition results further confirmed that January represents the peak period. Moreover, between 2015 and 2018, the search index for cough was higher compared to other years. The seasonal trend analysis reveals a close correlation between cough and the seasons, with cough reaching its peak during cold seasons. The results of six machine learning models and generalized additive models showed that meteorological factors had a typical delayed effect on coughing.