Abstract:
We consider a new method to improve the quality of training in gradient boosting as well as to increase its generalization performance based on the use of modified loss functions. In computational experiments, the possible applicability of this method to improve the quality of gradient boosting when solving various classification and regression problems on real data is shown.
Keywords:
gradient boosting, decision tree, loss function, machine learning, data analysis.
Presented by the member of Editorial Board:A. A. Lazarev
Citation:
N. S. Korolev, O. V. Senko, “Method for improving gradient boosting learning efficiency based on modified loss functions”, Avtomat. i Telemekh., 2022, no. 12, 78–88; Autom. Remote Control, 83:12 (2022), 1935–1943