Burst Detection-Based Selective Classifier Resetting

Concept drift detection algorithms have historically been faithful to the aged architecture of forcefully resetting base classifiers for each detected drift. This approach prevents underlying becoming outdated as distribution a data stream shifts from one concept another. In situations where both and temporal dependence are present within stream, forced can cause complications in classifier evaluation. Resetting too frequently when is performance appear successful, fact this misleading. research, novel architectural method determining resets, Burst Detection-Based Selective Classifier (BD-SCR), presented. BD-SCR statistically monitors changes determine if should be reset drifts. The experimental process compares predictive state-of-the-art detectors comparison “No-Change” detector using inform control decision. Results show that effectively reduces negative impact during through clear negation detector, but capable maintaining methods.