Two stage Decision Tree Learning from Multi-class Imbalanced Tweets for Knowledge Discovery

Abstract

Data Mining is an efficient technique for knowledge discovery from existing databases. The existing algorithms performance degrades when applied to the multi class imbalance dataset. The imbalance nature of twitter data set also hinders the process of efficient knowledge discovery. In this paper, we proposed an efficient learning approach for knowledge discovery from multi class imbalance datasets specifically designed for opinion mining. The proposed Under Sampled Imbalance Decision tree Learning (USIDL) approach uses decomposition of multi class into number of binary class samples followed by a unique technique for under sampling the instances from majority subset of each binary sample. The experimental results suggest that the proposed technique performs better than the existing C4.5 algorithm on six evaluation metrics.