AIExplainer
Machine Learning Intermediate 2 min read

What is stationarity?

A statistical property of a time series where the distribution of data remains constant over time

Stationarity means that the characteristics of a time series, such as its mean and variance, do not change significantly over time. This is important in many statistical and machine learning models, as they often assume that the data is stationary.

Imagine a river with a steady flow of water. If the river's flow rate and water level remain relatively constant over time, it can be considered stationary. However, if the flow rate or water level changes significantly due to factors like rainfall or drought, the river is no longer stationary.

A company's daily sales data may be considered stationary if the average sales and variability remain relatively constant over time. However, if the company introduces a new product or experiences a change in market trends, the sales data may no longer be stationary.

Stationarity is used in various fields, including finance, economics, and signal processing, to analyze and model time series data. It is often checked using statistical tests, such as the Augmented Dickey-Fuller test, to determine if a time series is stationary or not.

One common misconception is that stationarity requires the data to be perfectly constant, but in reality, small fluctuations are allowed as long as the overall distribution remains relatively stable.

The concept of stationarity has its roots in statistics and signal processing, dating back to the early 20th century. It has since become a crucial assumption in many machine learning and statistical models.

stability homogeneity invariance

Three products for different needs — explore what’s relevant to you.