An Adaptive Anomaly Threshold in Artificial Dendrite Cell Algorithm

The dendrite cell algorithm (DCA) relies on the multi-context antigen value (MCAV) to determine the abnormality of a record by comparing it with anomaly threshold. In practice, the threshold is pre-determined before mining based on previous information and the existing MCAV is inefficient when expose to extreme values. This causes the DCA fails to detect unlabeled data if the new pattern distinct from previous information and reduces the detection accuracy. This paper proposed an adaptive anomaly threshold for DCA using the statistical cumulative sum (CUSUM) with the aim to improve its detection capability. In the proposed approach, the MCAV were normalized with upper CUSUM and the new anomaly threshold was calculated during run time by considering the acceptance value and min MCAV. From the experiments towards 12 datasets, the new version of DCA generated a better detection result than its previous version in term of sensitivity, specificity, false detection rate, and accuracy.