面向群智感知车联网的异常数据检测算法

发布时间：2018-05-20 11:43

本文选题：车联网 + 群智感知　；参考：《湖南大学学报(自然科学版)》2017年08期

【摘要】：群智感知车联网利用普通用户的手机或平板电脑等智能终端获得交通数据,解决了车联网以低成本获取足够数据的问题,但却凸显了数据"质"的问题.为此,在分析群智感知车联网的数据结构及数据异常特点的基础上,提出一种适用于群智感知车联网的异常数据检测算法,并依此剔除异常数据,提高数据质量.算法利用核密度估计理论对车联网数据的概率密度进行估计,进而构建信任函数计算被检数据的信任度,后根据统计学理论将信任度小于0的数据判定为异常数据.最后对该算法的可行性及性能进行了仿真,结果表明该算法的性能可满足实用需求,且对比传统的统计检测法在检测率和误检率上具有更好的性能.
[Abstract]:Using intelligent terminals such as ordinary users' mobile phones or tablets to get traffic data, the group smart car network solves the problem of getting enough data at low cost, but it highlights the problem of data "quality". Based on the analysis of the data structure and the characteristics of data anomalies, an algorithm for detecting abnormal data is proposed, which can be used to eliminate abnormal data and improve the quality of data. The algorithm uses the kernel density estimation theory to estimate the probability density of the vehicle network data, and then constructs the trust function to calculate the trust degree of the tested data. Then, according to the statistical theory, the data whose trust degree is less than 0 are judged as abnormal data. Finally, the feasibility and performance of the algorithm are simulated, the results show that the performance of the algorithm can meet the practical needs, and it has better performance than the traditional statistical detection method in detection rate and false detection rate.
【作者单位】：福州大学物理与信息工程学院;
【基金】：国家自然科学基金海峡联合基金重点支持项目(U1405251);国家自然科学基金资助项目(61571129,61601126) 福建省自然科学基金资助项目(2015J01250,2016J01299)~~
【分类号】：U495

【相似文献】