R中时间序列数据的异常检测 [英] Outlier detection of time series data in R

问题描述

在R中检测业务销售数据中的异常值（意味着可能存在趋势和季节性）需要采取哪些步骤？

What are the steps needed to detect outliers in business sales data (which means there might be trends and seasonality) in R?

我了解了ACF，PACF ，残差，ARIMA模型（基本上是时间序列分析和建模）。我可以利用这些知识来帮助识别异常值吗？

I have learnt about ACF, PACF, residual, ARIMA model (basically, time series analysis and modelling). Can I use this knowledge to help me identify outliers?

是否还可以要求R指出哪个数据点是异常值？

Is it also possible to ask R to pinpoint which point of data is outlier?

非常感谢。

推荐答案

您可能会看到R中提供的以下软件包。

You may have a look at the following packages available in R.

R包 预测 使用时间序列的黄土分解来识别和替换离群值。

The R package forecast uses loess decomposition of time series to identify and replace outliers.

R包 tsoutliers 实现了 Chen and Liu 过程用于检测时间序列中的异常值。包装随附的文档中提供了有关过程和实现的说明。您可能还会看到此帖子。

The R package tsoutliers implements the Chen and Liu procedure for detection of outliers in time series. A description of the procedure and the implementation is given in the documentation attached to the package. You may also see this post.

这篇关于R中时间序列数据的异常检测的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！