Forecasting time series with 1-click machine learning inside the TICK stack

Plug and play, point and click — let’s predict the future!

Loud ML provides long-term predictive maintenance, so you’re always a step ahead. If you saw our previous tutorial on live data monitoring, you’re already aware that Loud ML is the first machine learning API natively integrated in the TICK stack, with more integrations on the way. Within Chronograf from InfluxData, machine learning can begin with a click of the mouse button. In fact, during setup, there’s absolutely no typing involved at all, saving frustrating typos and having to look up long strings of text.

In this tutorial, we’ll use the same data center monitoring for our sample data as our previous tutorial. Of course, any other live data that produces a regular output pattern will work, such as supply chain management and budgeting. The video includes the basics, while the text below contains more detail for reference. The video accompanying this tutorial uses live data.


  • You already have your own live data and that you’re familiar with the basics of Chronograf.
  • To run your own 1-click ML tasks, you will need the TICK stack add-on, which is open source. The add-on supports all standard InfluxDB features, so there’s no learning curve.

Let’s point and click using 1-click ML for time series forecasting

Since we’ve already trained our historical data in our previous tutorial, we’ll jump right in to forecasting.

Click on the Forecast button to display a calendar. Choose a start date in the past, as well as a target date any time in the future, and click on Apply. Here, we’ll choose the next hour.

Updates will roll in from the right, indicating your job is complete, and that you’re ready to look into the future through your favourite dashboard.

Now, find your existing dashboard and click on it to display the graph of the future points.

Change the time range to correspond with the forecast start time and date and the target time and date. Note that you can type these manually, or use a shortcut, such as ‘Next week’ or ‘Next month’, in the list available on the left of both calendars.

Click on Apply to display the data. A new, dark green curve — showing the expected pattern for future metrics — is displayed.

There’s a lot more being displayed in the graph above. First, there’s an anomaly, which is shown as the white vertical lines because in the previous tutorial, we turned on annotations.

More importantly for this tutorial, live data is arriving constantly, since our future predictions were set for just the next hour. The live data is displayed as the horizontal light green laser flashing up and down the screen. Live data is fast!

The good news is that the live data is producing the expected future pattern, which means the incoming data is considered normal, following the expected pattern. It also means that our forecast is pretty accurate — perfect!

It’s possible to train the same model again, with new data, at any time. Retraining might further improve accuracy. To do this, click on the Loud ML button on the Chronograf sidebar, then click on the Train drop-down to choose your retraining dates. Don’t forget, there are shortcuts on the left for faster selection.

When you no longer need a model, you can delete it with just a few clicks. Also from the Loud ML button in the Chronograf sidebar, click on the red Delete button, and then click on the green tick to confirm the request.

We’ve shown you the basics, and now you’re ready to create your first 1-click ML model within Chronograf. Download the TICK stack add-on via Docker, and visit to find everything else you need to start forecasting tomorrow, today!