通信基站关键指标数据异常检测与趋势预测的研究

Research on Anomaly Detection and Trend Prediction of Key Index Data of Communication Base Station

本文主要研究了智能运维中的异常检测与趋势预测的问题。采用高斯滤波器、归一化处理等方法,完成了对数据的预处理;利用Hampel滤波函数获取异常值,通过快速傅里叶变换计算异常周期,最后使用数据训练模型和决策树分类器,建立基于BP神经网络的滚动预测模型,实现时间序列中的异常预测和趋势预测。

This paper mainly studies the problems of anomaly detection and trend prediction in intelligent operation and maintenance. Gaussian filter, normalization processing and other methods are used to complete the preprocessing of the data;use the Hampel filter function to obtain abnormal values, calculate the abnormal cycle through fast Fourier transform, and finally use the data training model and decision tree classifier to establish The rolling prediction model based on BP neural network realizes the requirements of abnormal prediction and trend prediction.