Negative selection algorithm with variable-sized r-contiguous matching rule

We proposes a negative selection algorithm with variable-sized r-contiguous matching rule. The value of r can be used to balance between more generalization or more specification. A fixed r reaches a rough balance. We expect a variable-sized r to reach a refined balance. Experiments are performed to test the algorithm. Comparison to traditional negative selection algorithm and one-class SVM are conducted. The results demonstrate that the new scheme enhances the negative selection algorithm in efficiency for anomaly detection.