动态网络模糊域数据缺陷实时修正方法仿真

Simulation of Dynamic Network Fuzzy Domain Data Defect Real-Time Correction Method

缺陷数据大量存在于动态网络模糊域中,不仅严重影响了数据查询质量,还会扭曲数据分析结论,需要进行模糊域数据缺陷实时修正。传统的数据缺陷修正方法将网络模糊域缺陷数据协方差拟合代价函数转变为凸优化问题,通过对缺陷数据估计算子的更新实现数据修正。对于大规模数据缺陷修正问题,无法准确的估计数据缺陷情况,执行时间随着缺失比例的不断增加呈线性增长,无法满足实时性要求。提出一种基于遗传优化的动态网络模糊域数据缺陷实时修正方法。采用遗传优化方法估计模糊域缺陷数据的参数,获得最优数据参数,在最优参数基础上采用模糊域数据特有的多维度相关性对不同数据源间的近似度进行衡量,构建以模糊域缺陷数据源为中心的近似图,并将已填补的缺陷值作为观测值用于后续模糊域数据缺陷实时修正过程中,对缺陷数据缺失值进行顺序敏感的填补,完成模糊域数据缺陷实时修正。实验结果表明,所提方法加快了估计缺陷数据的参数的迭代速度,对于大规模数据缺陷修正问题,相比当前方法在执行时间和加速比性能上优势明显。

There are many defect data in the fuzzy network of dynamic network, which seriously affect quality of data query. Traditionally, modification methods of data defect are not impossible to accurately estimate the data de- fect situation and cannot meet the real - time requirements. A real - time modification method of data defect of fuzzy domain in dynamic network based on genetic optimization is put forward. The genetic optimization method is used to estimate parameters of defect data in fuzzy domain and obtain optimal data parameters. On the basis of optimal pa- rameters, the multidimensional interdependency of data in fuzzy domain is used to measure the approximation degree between different data sources and establish the approximate graph which takes defect data source in fuzzy domain as the center. Moreover, the filled defective value is regarded as the observed value which is used in the real - time cor- rection process of subsequent data defect in fuzzy domain. Finally, the missing value of defect data is sequentially and sensitively filled up, thus the real - time correction of data defect in fuzzy domain is completed. In simulation re- suits, the proposed method accelerates the iteration speed of estimating parameter of defect data. For large - scale da- ta defect correction, this method has obvious advantages in executive time and performance of speed - up ratio.