<p>Accurately predicting daily tourism demand is of great significance to resource allocation and market planning of tourist attractions. In this paper we combine group penalty with Poisson regressions to avoid multi-collinearity and over-fitting and improve prediction accuracy, propose three group penalized Poisson regressions to explore the significant factors influencing daily tourism demand, and develop a new group coordinate descent (GCD) algorithm to simultaneously complete group selection and group estimation. We also perform an empirical analysis using tourist data from Mount Siguniang, and show that GLASSO/GMCP/GSCAD penalized Poisson regressions outperform LASSO/Ridge/MCP/SCAD penalized Poisson regressions in terms of mean absolute error (MAE), root mean square error (RMSE) and symmetric mean absolute percentage error (SMAPE), where GMCP penalized Poisson regression exhibits the best predictive performance. Thus, GMCP penalized Poisson regression can more accurately capture the relationship between daily tourism demand and these predictive variables, where grouped variables such as date factors, climatic conditions, search engine analysis are the main factors influencing the daily tourism demand.</p>