基于广义回归神经网络的电网信息系统日志数据分析技术研究

摘要

In view of the grid system processing large volumes of logs data and slow processing speed in the current, based on existing data processing systems, and data analysis methods based on neural network, by using the Flume data acquisition system, the main structural model from HDFS and parallel programming mode MapReduce support integrated data calling program to optimize the traditional generalized neural network algorithm, and to improve the system for analyzing data to respond speed. Finally, select some samples of grid information system to carry out experimental testing: compared with other algorithms for data processing, GRNN data processing algorithm response time increased nearly 52%, to improve the accuracy of the results of about 10%.%针对目前电网信息系统处理日志数据量大且系统处理速度慢的问题,根据现有数据处理系统以及基于神经网络的数据分析方法,通过利用Flume数据采集系统、主从结构模型HDFS和并行编程模式的MapReduce综合支持的数据调用程序,优化传统的广义神经网络算法,提高系统分析数据做出响应的速度.最后选取电网信息系统的部分样本进行实验测试:较其他数据处理算法来说,GRNN算法数据处理响应时间提高近52%,处理结果的准确性提高10%左右.