文献类型：期刊文献

中文题名：大数据下网络重叠数据优化识别仿真

英文题名：Optimization Recognition Simulation of Network Overlapped Data Under Large Data

作者：赵芳[1]

第一作者：赵芳

机构：[1]新乡学院计算机与信息工程学院

第一机构：新乡学院计算机与信息工程学院

年份：2018

卷号：35

期号：7

起止页码：381-384

中文期刊名：计算机仿真

外文期刊名：Computer Simulation

收录：CSTPCD;;北大核心:【北大核心2017】；

语种：中文

中文关键词：大数据;网络;重叠数据;优化识别

外文关键词：Big data;Network;Overlapped data;Optimization recognition

摘要：对网络重叠数据的检测与排除,能够有效提高网络数据处理精度。大数据下对重叠数据的检测,需要计算数据聚类离散程度的期望,获得最佳聚类数目以确定聚类中心,完成重叠数据的检测,进而对其进行排除。传统方法结合量子滤波器,对数据点进行了一定程度的平滑,判断数据的真实度,但忽略了对数据的聚类中心的求取,导致重叠数据优化识别精度偏低。提出基于相关性分析和GAS算法的优化识别方法,将测量数据进行聚类,计算对应不同聚类数目的聚类离散程度的期望,和聚类数目对应的间隙统计量值,获得最佳聚类数目以此确定聚类中心,实现重叠数据检测;依据检测结果,对需要进行优化识别的重叠数据按照时间顺序,采用类型相同时段相同的负荷加权平均值法进行排除。实验证明,所提方法对重叠数据具有较好的优化识别效果,且提高了测量数据负荷预测的准确率。
In traditional methods, we often ignore the clustering center of data, which leads to the low accuracy of optimization and recognition for overlapped data. In this article, we propose the recognition and optimization method based on correlation analysis and GAS algorithm. This method clustered the measurement data, and then calculated the expectation of clustering dispersion degree corresponding to different number of clustering and gap statistics corre- sponding to clustering number, so as to obtain the best clustering number. Thus, the cluster center could be deter- mined to achieve the detection of overlapping data. According to detection results, using method of weighted mean value with the same type and time interval, we chronologically eliminated overlapping data which were needed to be i- dentified and optimized. Simulation results show that the proposed method has a good effect of optimization and recog- nition on overlapping data, and improves the accuracy of load prediction for measurement data.