【摘要】 针对已有实值否定选择算法检测器生成过程的不足,提出了一种优化的检测器生成算法。充分利用自体空间的分布,优化检测器生成的中心位置,扩大检测器的半径,尽可能生成覆盖范围大的检测器;使用覆盖率期望值作为算法结束的一个控制参数,有效地避免了冗余检测器的产生。建立了异常检测系统的形式化描述,定义了一个新的异常检测性能衡量指标——错误率。最后,通过人工合成数据集2DSyntheticData以及实际的Iris数据集及Bio-medical数据集对算法进行了验证。试验结果表明,相比V-detector算法,本文算法提高了检测率,降低了错误率,减少了所需检测器数量,整体检测性能较优。更多还原

【Abstract】 An optimized detector generation algorithm was proposed to solve the shortcomings of existing real-value negative selection algorithm.This algorithm fully utilizes self distribution,optimizes the center of the detector and expands the radius of the detector in order to generate a detector with large coverage.The expected coverage was used as one of the control parameters to end the algorithm so that it can effectively avoid the generation of redundant detector.Moreover,the anomaly detection system and it formal description were established.The error rate as a new index for anomaly detection performance was defined.The 2DSyntheticData and the actual Irish data set and biomedical data were used to test the algorithm.Experimental results show that,compared with V-detector,the algorithm improves detection rate,reduces the error rate and the required number of detectors.So it has a better detection performance.更多还原