[英]What is the most appropriate machine learning model to detect abrupt changepoints in time-series data?
I have a dataframe with a column for time in which anomalies occur.我有一个 dataframe 列,其中有一个异常发生时间的列。 The anomalies are of a sudden changepoint in the time column.异常是时间列中的突然变化点。
The changepoint is the blue point in the graph.变化点是图中的蓝点。
My goal is to identify these points as anomalies and mark them.我的目标是将这些点识别为异常并标记它们。
I've tried searching for anomaly detection ML models for this problem such as:我已经尝试为这个问题搜索异常检测 ML 模型,例如:
None of these ML models succeed at identifying these anomalous points but maybe i missed something.这些 ML 模型都没有成功识别这些异常点,但也许我错过了一些东西。
Any help will be much appreciated.任何帮助都感激不尽。
With an anomaly detection algorithm like OneClassSVM, GaussianMixtureModel, IsolationForest etc your training setup and features will be the deciding factor in success or not.使用 OneClassSVM、GaussianMixtureModel、IsolationForest 等异常检测算法,您的训练设置和特征将成为成功与否的决定因素。
The kind of anomaly shown there will be easily detected if you transform your data to differences or rate-of-change between consecutive time-windows.如果您将数据转换为连续时间窗口之间的差异或变化率,将很容易检测到那里显示的异常类型。
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.