PSO优化BP神经网络入侵检测模型

Intrusion Detection Model Based on PSO Optimized BP Neural Network

文中提出了一种基于变惯性因子粒子群优化的BP网络学习算法。该算法用PSO算法代替了传统的BP算法,克服了BP算法易陷入局部最小值的不足,并且将该算法应用于入侵检测中。在预处理数据时,采用了信息增益的方法,提取出含信息量多的特征作为BP网络的输入向量。通过实验仿真比较,证明了该算法的收敛速度快,迭代次数少,准确率较高。

This paper proposes a BP networks learning algorithm based on changed inertia particle swam optimization algorithm. This algorithm substitutes the traditional BP algorithm, overcomes the problem that the traditional BP algorithm would be easy to fall into the local minima, and thus is applied to intrusion detection. Information gain is used in data preprocessing, the characteristics with much information are extracted as the input vectors of BP network. Simulation and comparison indicate that the BP networks based on changed inertia particle swam optimization algorithm is of faster convergence rate, less iterations, and higher accuracy.