基于Haar小波的队列动态多尺度自适应智能预测

Adaptive intelligent prediction of queue dynamics based on Haar wavelet multiscale fusion

为降低网络拥塞风险和实现对网络队列缓存资源状态预测,基于智能控制技术和多尺度表示方法,通过引进自适应补偿模糊因子,运用Haar小波的优良特性,构造了一类在线无监督学习实时预测补偿模糊神经网络,提出了基于Haar队列动态多尺度融合自适应智能预测方案.仿真表明,该预测策略对于自相似特性数据流在队列缓存动态具有较好的预测能力.

By introducing an adaptive compensated fuzzy factor, an online non-supervised learning real-time prediction compensated fuzzy-neural network （OnSLRPCFNN） is designed. A novel strategy of adaptive intelligent prediction of queue length based on Haar wavelet multiscale fusion of queue dynamics is developed, which can decrease risk of network self-similar data flow congestion and improve the capability for the prediction of queue dynamics. The simulation shows the effectiveness of the proposed approach.