离群值
异常检测
计算机科学
稳健性(进化)
数据挖掘
恒虚警率
相关性
大数据
人工智能
模式识别(心理学)
数学
生物化学
化学
几何学
基因
作者
M. Veera Brahmam,S. Gopikrishnan
标识
DOI:10.1093/comjnl/bxad034
摘要
Abstract An outlier in the Internet of Things is an immediate change in data induced by a significant difference in the atmosphere (Event) or sensor malfunction (Error). Outliers in the data cause the decision-maker to make incorrect judgments about data analysis. Hence it is essential to detect outliers in any discipline. The detection of outliers becomes the most crucial task to improve sensor data quality and ensure accuracy, reliability and robustness. In this research, a novel outlier detection technique based on spatial, temporal correlations and attribute correlations is proposed to detect outliers (both Errors and Events). This research uses a correlation measure in the temporal correlation algorithm to determine outliers and the spatial correlation algorithm to classify the outliers, whether the outliers are events or errors. This research uses optimal clusters to improve network lifetime, and malicious nodes were also detected based on spatial–temporal correlations and attribute correlations in these clusters. The experimental results proved that the proposed method in this research outperforms some other models in terms of accuracy against the percentage of outliers infused and detection rate against the false alarm rate.
科研通智能强力驱动
Strongly Powered by AbleSci AI