【作者】 李娜娜；

【导师】 顾军华；

【作者基本信息】 河北工业大学 ， 计算机应用技术， 2005， 硕士

【摘要】 生物免疫系统作为一种新的智能信息处理系统,具有免疫识别、免疫记忆、自适应等许多优良的特性,有着非常重要的理论研究价值和应用前景。基于免疫的异常检测是国内外学术界研究的热点和难点问题。本文主要针对Kim提出的人工免疫检测模型进行了研究,取得了一定的成果。 首先,综合分析了Kim提出的人工免疫模型检测异常方法的特点和存在的问题,提出了一种新的基于免疫的层次异常检测模型,并进行了符号化处理。模型通过免疫过程中的分层处理能够逐层逐级的、分门别类的检测各类异常。 其次,提出上述模型也可以作为免疫算法的一种改进。新算法在免疫遗传算法的基础上引入了负选择机制和层次思想。它通过负选择避免了“早熟”现象,通过分层技术又可以逐层分类检测出各式的异常。实验证明,新方法收敛速度较快。 再次,应用聚类方法,建立了算法的评价指标,通过比较个体的分布规律来评价算法的效率。 最后,通过对求解多峰值函数的全局最值问题和求多峰值函数的所有局部峰值的问题的结果分析比较,验证了新模型和方法的有效性。更多还原

【Abstract】 As a new intellective and information processing system, biological immune system has many excellent characteristics such as immune recognition, immune memory, self-adaptability and so on. So its research is great important. Recently anomaly detection based on immunology as a new method becomes very hot. In this paper, we research and improv the primal artificial immune model which Kim proposed, tod attaind some achievements.First, we analyze the characteristic and limitation of Kim artificial immune model. Then we introduce a multi-layer idea, and give a hierarchical model of anomaly detection based on immunology which is proposed first here. Futhermore we use mathematical symbol to abstract this new model. And this model can detect various anomaly layer by layer in order to improve the efficiency of detection.Second, we propose that mis new model may be looked as an improvement of immune algorithm. Immune algorithm is superior genetic algorithm by variety and memory mechanism, while the improvement of immune algorithm is superior immune algorithm by negative selection and hierarchical idea. This new algorithm can deal with the problem of precocity due to negative selection, and it also can detect various anomaly due to layered technology.Third, we propose a kind of method that evaluate the quality of algorithm by introducing index of clustering as the basis of evaluation.At last, we prove the validity of new method by experiment. And we mainly do two kind of experiment: one is to solve max. or min. of multi-peak-value function; the other is to solve all local extrema of multi-peak-value function.更多还原