Skip to content

Identify trends and seasonality over time

Time Series data consist of data points indexed by a DateTime feature, which can be visualised with a Line Plot.

Setting up your time series data

Once you load your dataset with at least one feature of DateTime type, you can visualise Numerical values on the Line Plot tab.

If you're using an example dataset, click on restaurant-sales.csv from the Datasets tab, which shows the aggregated monthly retail sales amount in the U.S. from restaurants and similar establishments, in USD millions.

By default, only features of DateTime type are displayed as options for the X-Axis, and Numerical features on the Y-Axis.

Zooming into specific time windows

The Time Range Filter feature allows you to specify the window of time to display the line plot for by sliding the circles.

Note that the start and end DateTime are displayed in the ISO 8601 format.

Time series data can be decomposed into multiple components.

  • A trend is an overall pattern in a given period of time
  • Seasonality refers to cyclic behaviours which happen at regular intervals

In the case of this restaurant sales data, for example, one can see

  • Uptrend: More people are spending more dollars over time
  • Monthly seasonality: Spending increases during warmer times of the year, such as spring and summer months, and during holiday seasons.

From this line plot, you can also observe anomaly in the beginning of 2020, marked by a sudden decrease in overall spending at restaurants as COVID-19 hit the world.

restraunt_sales