Research on Alarm Cleaning and Outlier Detection Algorithms of Network Data Analysis

With the rapid development of 5G network, network architecture is becoming increasingly complex, and the number of network devices are also increasing significantly. Complex and diverse network data is generated during the daily operation of network devices. In addition, network alarms seriously threaten the network operation, and network outliers clearly indicate the existence of hidden dangers in the network. In order to guarantee effective network operation, this paper investigates two algorithms about alarm cleaning and outlier detection. An alarm cleaning algorithm is proposed and based on network function association relationship, which can effectively filter out the redundant alarm and facilitate intelligent alarm analysis. In addition, an outlier detection algorithm is proposed and based on the standard interquartile range (IQR) method and the outlier coefficient of data temporal correlation, which is helpful in detecting network outliers more effectively and facilitate network performance analysis and automatic operation and maintenance.