Contents
How to Detect anomalies in time series data?
The entire process of Anomaly Detection for a time-series takes place across 3 steps:
- Decompose the time-series into the underlying variables; Trend, Seasonality, Residue.
- Create upper and lower thresholds with some threshold value.
- Identify the data points which are outside the thresholds as anomalies.
What is anomaly in time series?
These observations are often referred to as anomalies. The anomaly detection problem for time series is usually formulated as identifying outlier data points relative to some norm or usual signal.
How to feed time series data to autoencoder?
Here, we’ll have a look at how to feed Time Series data to an Autoencoder. We’ll use a couple of LSTM layers (hence the LSTM Autoencoder) to capture the temporal dependencies of the data. To classify a sequence as normal or an anomaly, we’ll pick a threshold above which a heartbeat is considered abnormal.
How are autoencoders used in LSTM anomaly detection?
The general Autoencoder architecture consists of two components. An Encoder that compresses the input and a Decoder that tries to reconstruct it. We’ll use the LSTM Autoencoder from this GitHub repo with some small tweaks. Our model’s job is to reconstruct Time Series data. Let’s start with the Encoder:
How to use VAE for time series Anomaly Detection?
To do the automatic time window isolation we need a time series anomaly detection machine learning model. The goal of this post is to introduce a probabilistic neural network (VAE) as a time series machine learning model and explore its use in the area of anomaly detection.
When to use autoencoder for contextual anomaly?
It is in your interest to automatically isolate a time window for a single KPI whose behavior deviates from normal behavior (contextual anomaly – for the definition refer to this post ). When you have the problematic time window at hand you can further explore the values of that KPI.