Active Learning for Multiclass Cost-Sensitive Classification Using Probabilistic Models

Multiclass cost-sensitive active learning is a relatively new problem. In this paper, we derive the maximum expected cost and cost-weighted minimum margin strategies for multiclass cost-sensitive active learning. The two strategies can be viewed as extended versions of the classical cost-insensitive active learning strategies. The experimental results demonstrate that the derived strategies are promising for cost-sensitive active learning. In particular, the cost-sensitive strategies out-perform cost-insensitive ones on many benchmark data-sets and justify that an appropriate consideration of the cost information is important for solving cost-sensitive active learning problems.