基于Spark框架和PSO优化算法的电力通信网络安全态势预测

发布时间：2018-05-04 18:37

本文选题：Spark计算框架 + 粒子群算法　；参考：《计算机科学》2017年S1期

【摘要】：随着电力通信网络规模的不断扩大,电力通信网络不间断地产生海量通信数据。同时,对通信网络的攻击手段也在不断进化,给电力通信网络的安全造成极大威胁。针对以上问题,结合Spark大数据计算框架和PSO优化神经网络算法的优点,提出基于Spark内存计算框架的并行PSO优化神经网络算法对电力通信网络的安全态势进行预测。本研究首先引入Spark计算框架,Spark框架具有内存计算以及准实时处理的特点,符合电力通信大数据处理的要求。然后提出PSO优化算法对神经网络的权值进行修正,以增加神经网络的学习效率和准确性。之后结合RDD的并行特点,提出了一种并行PSO优化神经网络算法。最后通过实验比较可以看出,基于Spark框架的PSO优化神经网络算法的准确度高,且相较于传统基于Hadoop的预测方法在处理速度上有显著提高。
[Abstract]:With the continuous expansion of power communication network scale, power communication network continuously produces massive communication data. At the same time, the attack means to the communication network is evolving, which poses a great threat to the security of the power communication network. In view of the above problems, combining the advantages of Spark big data computing framework and PSO optimization neural network algorithm, a parallel PSO optimization neural network algorithm based on Spark memory computing framework is proposed to predict the security situation of power communication network. In this paper, firstly, Spark framework is introduced, which has the characteristics of memory computing and quasi-real-time processing, which meets the requirement of big data processing in power communication. Then PSO optimization algorithm is proposed to modify the weights of neural networks to increase the learning efficiency and accuracy of neural networks. Then a parallel PSO optimization neural network algorithm is proposed according to the parallelism of RDD. Finally, the experimental results show that the accuracy of PSO optimization neural network algorithm based on Spark framework is high, and compared with the traditional prediction method based on Hadoop, the processing speed is improved significantly.
【作者单位】：中央财经大学信息学院;北京国电通网络技术有限公司;人力资源和社会保障部人事考试中心;
【基金】：国网科技部项目(SGTYHT/14-JS-188)资助
【分类号】：TM73;TN915.853;TP393.08

【相似文献】