基于块状bootstrap技术的Bagging Trees集成算法研究

集成算法已经成为机器学习研究的一大热点,已有许多改进的集成算法,但对病态数据的集成研究并不常见。本文通过对一海藻繁殖案例的研究,提出了一种基于块状bootstrap技术的集成算法,并将其与几种常用的集成算法比较研究得出,在对于一些病态数据而言,该算法往往比其它算法具有更小的模型推广误差和更高的预测精度的优点。

Study on Bagging Trees Integration Algorithm Based on Block Bootstrap Technology

At present,integration algorithm is popular in the field of machine learning,and there are many people proposing to improve algorithms on it,but there are few integration algorithms on Ill-conditioned data.This paper introduces a new bagging trees integration algorithm based on block bootstrap technology by a case study of one kind of algae's reproduction.Compared with several popular integration algorithm,it often has smaller generalization error and higher prediction precision when it deals with Ill-conditioned data.