A concept drift-tolerant case-base editing technique

Abstract

The evolving nature and accumulating volume of real-world data inevitably give rise to the so-called "concept drift" issue, causing many deployed Case-Based Reasoning (CBR) systems to require additional maintenance procedures. In Case-base Maintenance (CBM), case-base editing strategies to revise the case-base have proven to be effective instance selection approaches for handling concept drift. Motivated by current issues related to CBR techniques in handling concept drift, we present a two-stage case-base editing technique. In Stage 1, we propose a Noise-Enhanced Fast Context Switch (NEFCS) algorithm which targets the removal of noise in a dynamic environment, and in Stage 2, we develop an innovative Stepwise Redundancy Removal (SRR) algorithm which reduces the size of the case-base by eliminating redundancies while preserving the case-base coverage. Experimental evaluations on several public real-world datasets show that our case-base editing technique significantly improves accuracy compared to other case-base editing approaches on concept drift tasks, while preserving its effectiveness on static tasks.

Concept drift region identification via competence-based discrepancy distribution estimation

“…By utilizing the competence model as a space partition technique, competence-based drift detection method [6] highlight a drift region of the problem space through case-base competence by TPCA [10]. However it ignores the fact that space partitions may overlap one another, and only focuses on competence areas that have a large discrepancy.…”

“…Their competence-based drift detection method also quantitatively describes when, how and where data change takes place, and demonstrates good performance in different scenarios [6]. The method they used to highlight drift-affected regions is based on competence models, called Top-P-Competence Areas (TPCA) [10]. However, TPCA only focuses on competence areas that have a large discrepancy.…”

“…SEA Concepts [17]: Data points are uniformly randomly sampled from space [0, 10] 3 . A data point will be assigned to class 1 if the sum of the first two feature values is not larger than a threshold θ, otherwise it will be assigned to class 0.…”

“…One is Top-P-Competence Areas (TPCA) [10], and the other is Kulldorff Spatial Scan Statistic (KSSS) [5]. In TPCA, parameter p indicates the confidence of the identified areas.…”

“…Little research endeavor has been made to explain how and where the data changes. Lu et al [6], [10] and Dasu et al [5] mention that their methods can locate the regions in which the data changes; however one issue should be addressed is that, the drift regions they identify are based on how their data model partitions data feature space. If the partitions are too coarse, the result will be imprecise.…”