相空间重构与改进SMA优化SVR的网络流量预测

Network traffic prediction combined phases space reconstruction with improved slime mould algorithm optimizing support vector regression

为提高网络流量预测精度,提出结合相空间重构与改进黏菌优化支持向量回归的预测模型。为解决黏菌算法收敛慢、易得局部最优的不足,引入3种形态对立学习对种群进行初始化,提高种群多样性;利用非线性反馈因子更新机制,均衡全局搜索与局部开发;设计柯西-高斯混合变异对最优解变异,扩展搜索空间,避免陷入局部最优。利用改进黏菌算法对支持向量回归优化调参,有效解决超参初值敏感缺陷,提高学习精度和收敛速度,以此构建网络流量预测模型。实验结果表明,改进模型预测误差更小,能够实现高精度和实时性预测要求。

For promoting the prediction accuracy of network traffic,a prediction model based on improved slime mold algorithm and optimized support vector regression combined with phases space reconstruction was proposed.To solve the problems of slow convergence and easiness to fall into local optimization of slime mold algorithm,three forms of opposite learning were introduced to initialize the population and improve the diversity of the population.A nonlinear feedback factor updating was introduced to balance the global search and local development.The Cauchy-Gaussian mixture mutation was designed to perturb the optimal solution and expand the search space to avoid the algorithm falling into local optimization.The improved slime mold algorithm was used to optimize the factors of support vector regression,which improved the learning accuracy and convergence speed by effectively solving the defects that the traditional parameter adjust method in the support vector regression is easy to get the minimum solutions and it is sensitive to initial parameters values,and a new network traffic prediction model was constructed.The results show that the improved model has less prediction error,and it can meet the high accuracy and real-time requirements.