高斯过程机器学习的基线计算算法

Baseline Algorithm Based on Gaussian Process Machine Learning

为了对网络监视领域中样本进行预测和相关处理,大多数研究在计算基线时都忽略了样本的概率特征,未能结合样本的数据分布,对样本进行相关的处理,忽略了利用样本的周期特性和数据分布对样本进行相关处理的改进空间.因此,本文分析样本历史数据的噪音,通过引入高斯过程机器学习方法,提出基于周期样本的高斯过程机器学习方法,通过采用复合核函数,实现了网络主动监控中的基线计算.首先对"周期数据"进行聚类处理,同时将核函数拆分为全局核函数部分和局部核函数部分,使用聚类点训练全局核函数部分;使用局部点训练局部核函数.通过实验,与其它算法相比大大提高了效率,而且保证了近似的准确性.最终保障网络安全、提升网络性能和用户满意度.

The baseline calculation is an important issue in the field of network monitoring. As to deal with the data, most researches just ignore the probability characteristics of the data, which fails to combine data distribution to predict the data and make the related processing and loses room for improvement in this field. Through the use of the compound kernel functions, the baseline is calculated in the proactive monitoring network. First, do clustering process for cycle data. Then we split the kernel functions for the global kernel functions and the local kernel functions, with using the cluster point to train the global kernel functions and the local points to the local kernel functions. Therefore, this article analyses the historical data＇s noise first, then make the prediction with the Gaussian process ma- chine learning. The experiment, compared with other algorithms, shows improvement in the field of efficiency and accuracy.