Evaluation of the ability of the random forest algorithm in machine learning for studying construction hydraulics

Tóm tắt

The Decision Tree and Random Forest algorithm is a "black box" prediction model, this algorithm is formed based on the "binary tree" structure. The study conducted an analysis of the structure of the Random Forest algorithm, built a process for analyzing and predicting a hydraulic factor using the regression algorithm. In particular, Pi theory is used to analyze and determine the objective function, thereby determining the input data fields for the Machine Learning model, coordinating experimental data with physical experimental model of the hydraulic jump in the trapezoidal channel. The study analyzed machine learning models according to Decision Tree and Random Forest algorithms, the research results showed good computational efficiency, strong correlation coefficient (R2 ≥ 0.9), other statistical indicators are very close to the ideal point (zero), the MAPE is from 3% to 6%. The study also shows that the Random Forest model has better prediction performance than the Decision Tree for the hydraulic factors of water jumping in an horizontal trapezoidal channel.