Robust classification systems for imprecise environments

摘要: In real-world environments it is usually difficult to specify target operating conditions precisely. This uncertainty makes building robust classification systems problematic. We show that possible build a hybrid classifier will perform at least as well the best available for any conditions. performance extends across wide variety of comparison frameworks, including optimization metrics such accuracy, expected cost, lift, precision, recall, and workforce utilization. some cases, can actually surpass known classifier. The also efficient build, store, update. Finally, we provide empirical evidence needed many problems.