Time Series Properties

Understand the main properties of time series data, such as frequency, range and trend.

We'll cover the following


Frequency refers to the level of granularity of the data—for example, is it one observation per hour, per day, per week? It's one of the most important properties regarding time series data, and it also has a direct link with the forecasting task we have.

Imagine we have weekly sales data but need to make daily forecasts. How would we do it? What if we have daily rain level data but need to forecast rain weekly?

Going from high granularity to a lower one is usually easier, but the way we aggregate values will depend on what kind of data we have. Here are some of the most common examples:

  • From daily sales to weekly sales: We could just sum sales for every day in a week to get the sales for that week.

  • From daily closing prices of stock to weekly closing prices of stock: Since the closing price is the last price available, we should use, for each week, the closing price of the last day of that week.

  • From daily minimum temperatures to weekly minimum temperatures: We get the minimum temperature out of all the minimum temperatures for days in that week.

Get hands-on with 1200+ tech skills courses.