Vector AutoRegression — The Multivariate Time Series Forecasting Method We Must Know

Image: source

In our previous blog, we’ve learnt the basic of multivariate time series and we used XGBoost regression method to forecast a dependent variable.

In this blog, we’ll see how to use VAR model for multivariate time series analysis.

For, univariate time series, please refer my blog — 10 Time Series Forecasting Methods We Should Know.

Vector AutoRegression (VAR)

Vector autoregression (VAR) is a statistical model used to capture the relationship between multiple quantities as they change over time. — reference

This model assumes that the passed time series dataset is stationary; that means, dataset doesn’t have any trend or seasonality. In case of non-stationary data, we need to transform it to a stationary dataset.

Datasets

We’ll use the same dependent variable: UK Industry and Services — Sales data and independent datasets: UK Consumer Price Index, UK Inflation Rate, UK Real Household Gross Disposable Income Per Capita, UK Population as we used in our previous blog. Refer the Datasets section at-Multivariate Time Series Forecasting using XGBoost.