What is Time Series Forecasting?

Time series forecasting is exactly what it sounds like; predicting unknown values. Time series forecasting involves the collection of historical data, preparing it for algorithms to consume, and then predicting the future values based on patterns learned from the historical data.

There are numerous reasons why companies may be interested in forecasting future values, namely GDP, monthly sales, inventory, unemployment, and global temperatures:

• A retailer may be interested in predicting future sales at an SKU (stock keeping unit) level for planning and budgeting.
• A small merchant may be interested in forecasting sales by store, so it can schedule the right resources (more people during busy periods and vice versa).
• A software giant like Google may be interested in knowing the busiest hour of the day or busiest day of the week so that they can schedule server resources accordingly.
• The health department may be interested in predicting the cumulative COVID vaccinations administered so that they can further predict when herd immunity is expected to kick in.

Type of Time Series Forecasting

There are three types of time series forecasting. Which one you should use depends on the type of data you are dealing with and the use-case in hand:

Univariate Forecast

A univariate time series, as the name suggests, is a series with a single time-dependent variable. For example, if you are tracking hourly temperature values for a given region and want to forecast the future temperature using historical temperatures, this is univariate time series forecasting. Your data may look like this:

Multivariate Forecast

On the other hand, a Multivariate time series has more than one time-dependent variable. Each variable depends not only on its past values but also has some dependency on other variables. This dependency is used for forecasting future values.

Consider the above example and suppose that our dataset includes other weather-related attributes over the same time period, such as perspiration percent, dew point, wind speed, etc., along with the temperature values. In this case, there are multiple variables to be considered to optimally predict temperature. A series like this would fall under the category of multivariate time series. Your dataset will look like this now:

You are still forecasting temperature values for the future but now you can use other available information in your forecast as we assume temperature values will be dependent on these factors as well.

Image Source: Van Nguyen

When we are dealing with multivariate time series forecasting, the input variables can be of two types:

• Exogenous: Input variables that are not influenced by other input variables and on which the output variable depends.
• Endogenous: Input variables that are influenced by other input variables and on which the output variable depends.