论文部分内容阅读
集成学习是一种受到广泛认可和使用的机器学习算法.为此提出一种新的多类集成学习算法,即AdaBoost belief.此算法改进多类集成学习算法AdaBoost·SAMME,使每个基分类器对于每个类别都有权重信息.这种类别上的权重被称为类别信念,可通过计算每次迭代中各个类别的正确率得到.将所提出的算法与原有的AdaBoost·SAMME算法从预测准确率、泛化能力以及理论支持等方面进行比较发现:在高斯数据集、多种UCI数据集以及基于日志的多类别入侵检测应用中,该算法不但具有更高的预测准确率和泛化能力,而且当类别数目增加,即类别更难以预测时,其分类错误率较原有AdaBoost·SAMME算法上升得更缓慢.
Integrated learning is a widely accepted and used machine learning algorithm. To this end, a new multi-class integrated learning algorithm is proposed, AdaBoost belief. This algorithm improves a variety of integrated learning algorithms AdaBoost · SAMME, so that each base classifier Weights are weighted for each category.The weights on this category are called category beliefs and can be obtained by calculating the correctness of each category in each iteration.The proposed algorithm is compared with the original AdaBoost • SAMME algorithm Accuracy, generalization ability and theoretical support. The results show that this algorithm not only has higher prediction accuracy and generalization ability in Gaussian datasets, multiple UCI datasets and log-based multi-category intrusion detection applications, , And its classification error rate is slower than the original AdaBoost • SAMME algorithm when the number of categories increases, that is, the categories are more unpredictable.