AIExplainer
Machine Learning Intermediate 2 min read

What is nonstationarity?

A statistical concept where a process or system's properties change over time or space

Nonstationarity refers to the idea that a system or process does not have a fixed set of statistical properties, such as mean or variance, that remain constant over time or space. Instead, these properties change, making it challenging to model or predict the system's behavior.

Imagine trying to predict the water level in a river. If the river's water level remains relatively constant, it's like a stationary system. But if the water level changes significantly due to factors like rainfall or tidal patterns, it's like a nonstationary system, making predictions more difficult.

Financial markets, weather patterns, and traffic flow are all examples of nonstationary systems, where properties like volatility, temperature, or traffic volume change over time.

Nonstationarity is often encountered in time series analysis, signal processing, and machine learning, where it can affect the accuracy of models and predictions. Techniques like differencing, normalization, or using nonstationary models can help address nonstationarity.

One common misconception is that nonstationarity is the same as randomness or noise. While related, nonstationarity refers specifically to changes in statistical properties over time or space, not just random fluctuations.

The concept of nonstationarity has its roots in statistics and signal processing, dating back to the early 20th century. It has since become a crucial aspect of many fields, including machine learning, economics, and environmental science.

non-stationarity time-varying space-varying dynamic

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