| 摘要: |
| 针对网络流量预测准确率不够高的问题，结合当下流行的集成学习（Ensemble Learning），提出一种Stacking集成多种模型的网络流量预测方法；将天气因素量化后作为输入，使用7个机器学习模型分别对网络流量进行预测，然后根据对不同模型预测结果的Pearson相关系数的分析，选取相关性较弱的5个模型作为Stacking的基模型，进行网络流量的预测，并与不考虑天气因素的预测结果进行比较；结果显示：Stacking方法相较于各基模型都有更好的表现，同时，天气因素的加入使得模型预测结果的准确性提高了；Stacking方法将不同的预测方法进行组合，相较于神经网络方法能以不同模型对数据进行不同角度的处理，能获得比一般方法准确率更高的预测结果，对于网络流量的预测具有一定的实用价值。 |
| 关键词: 流量预测 多模型 机器学习 Stacking |
| DOI： |
| 分类号: |
| 基金项目: |
|
| Network Traffic Prediction Based on Stacking Integration Model |
|
ZHU Guo-sen,ZHENG Xiao-liang
|
|
School of Electrical and Information Engineering，Anhui University of Science and Technology,Anhui Huainan 232000,China
|
| Abstract: |
| Aiming at the problem that the accuracy of network traffic prediction is not high enough，a network traffic prediction method integrating multiple models is put forward in combination with currently popular Ensemble Learning.The weather factors are quantified as input，and 7 machine learning models are used to predict the network traffic respectively.Then，based on the analysis of the Pearson correlation coefficients of the prediction results of different models，5 models with weak correlation are selected as the basic model of stacking to predict network traffic and compare it with predictions that do not consider weather factors.The results show that the stacking method has better performance than the basic models.At the same time，the addition of weather factors makes the accuracy of the model’s prediction results improved. Compared with the neural network method，the Stacking method combines different prediction methods,the data can be processed from different angles with each basic model，and the prediction results are more accurate than the general method.It has certain practical value for the prediction of network traffic. |
| Key words: traffic forecast multiple model machine learning Stacking |
