Clarify the Importance of Tree Numbers in Random Forest Regression

• In Event: AI, Machine Learning, and Data Science in Modeling and Prediction: New Strategies and Applications

Abstract

Random forest regression has been increasingly adopted in educational research for both predictive modeling and identifying key predictors of outcome variables. However, only a minority of studies report critical model parameters or describe their tuning procedures. Addressing this methodological gap, the present study systematically investigated the impact of the ntree parameter, the number of trees in a forest, on predictive accuracy and variable importance rankings. Our results revealed that even when applied to the same dataset, the minimum ntree required varies substantially depending on the specific analytical objective. We conclude with task-specific recommendations for parameter selection and emphasize the necessity of reporting the ntree parameter when applying random forest regression.

Authors

• YaShan Zhuang, East China Normal University

• Yi Jiang, East China Normal University