Augmented Query Strategies for Active Learning in Stream Data Mining

Abstract

Active learning is used in situations where the amount of unlabeled data is abundant but it is costly to manually label the data. So, depending on our available budget, from all unlabeled instances we are to select only a subset of them to ask the oracle for manual labeling. Thus, the query strategy, i.e., how relevant instances are selected to be sent to the oracle, plays an important role in active learning. Though active learning is a very established research area, only a few research works have been done on it in the context of stream data mining. Active learning for stream data is more challenging than for static data because the repetition of queries is not feasible as revisiting the data is almost impossible. In this paper, we propose two augmented query strategies for active learning in stream data mining, namely, Margin Sampling with Variable Uncertainty (MSVU) and Entropy Sampling with Uncertainty using Randomization (ESUR). These two strategies are derived and improved from the existing methods of Variable Uncertainty (VU) and Uncertainty using Randomization (UR) respectively. We evaluate the effectiveness of our proposed MSVU and ESUR strategies by comparing them against the original VU and UR on 6 different datasets using two base classifiers: Leveraging Bagging (LB) and Single Classifier Drift (SCD). Experimental results show that our proposed strategies offer promising outcomes for various datasets and detecting concept drift in the data.