Global and Local (Glocal) Bagging Approach for Classifying Noisy Dataset
Peng Zhang, Zhang Zhiwang, Aihua Li, Yong Shi · International Journal of Software and Informatics · 2008
Learning from noisy data is a challenging task for data mining research. In this paper, we argue that for noisy data both global bagging strategy and local bagging strategy suer from their own inherent disadvantages and thus cannot form accurate predic- tion models. Consequently, we present a Global and Local Bagging (called Glocal Bagging: GB) approach to tackle this problem. GB assigns weight values to the base classiers under the consideration that: (1) for each test instance Ix, GB prefers bags close to Ix, which is the nature of the local learning strategy; (2) for base classiers, GB assigns larger weight values to the ones with higher accuracy on the out-of-bag, which is the nature of the global learning strategy. Combining (1) and (2), GB assign large weight values to the classiers which are close to the current test instance Ix and have high out-of-bag accuracy. The diversity/accuracy analysis on synthetic datasets shows that GB improves the classier en- semble's performance by increasing its base classier's accuracy. Moreover, the bias/variance analysis also shows that GB's accuracy improvement mainly comes from the reduction of the bias error. Experiment results on 25 UCI benchmark datasets show that when the datasets are noisy, GB is superior to other former proposed bagging methods such as the classical bagging, bragging, nice bagging, trimmed bagging and lazy bagging.