Affiliations: [a] Key Lab of Intelligent Information Processing of Chinese Academy of Sciences, Institute of Computing Technology, CAS, Beijing, China | [b] University of Chinese Academy of Sciences, Beijing, China | [c] Central Software Institute, Huawei Technologies Co. Ltd., Beijing, China
Correspondence:
[*]
Corresponding author: Haocheng Wang, Institute of Computing Technology, Chinese Academy of Sciences, Beijing 100190, China. E-mail:hossenwang@126.com
Abstract: Ensemble learning via manipulating the training set is an effective technique for improving classification accuracy. In this work, we investigate the strategy how to combine learning set resampling method and random subspace method applied in high-dimensional domains. We propose a new procedure, Bag of Little Bootstraps on Features (BLBF), which works by combining the results of bootstrapping multiple feature subsets of the original dataset using the random subspace method. Our empirical experiments on various high-dimensional datasets demonstrate that our proposed approach outperforms the state-of-the-art instance-based resampling learning algorithm BLB and its two relevant variants, in terms of classification performance. In addition, we also investigate the effect of hyperparameters on classification performance, which shows that the parameters can be easily set while maintaining a good performance.
Keywords: Ensemble learning, bag of little bootstraps on features, high-dimensional data, classification
