Skip to content
 
Try for free

ThingsBoard Cloud

Choose your data region

Your data stays in the region you choose, for residency and compliance. No credit card required.

North America Data stored in North America Europe Data stored in the European Union

Rather run it yourself? Install on your own servers

IoT Time Series Forecasting Models

A prediction model in Trendz is a no-code time series forecasting model for IoT telemetry. It learns from the history of every device or asset in a profile and predicts what comes next: future values, limit crossings, period totals, or target progress. Trendz picks the forecasting method for each item, so no machine learning expertise is needed, and you can bring your own Python model when you want to. Predictions return to ThingsBoard as telemetry and alarms, so dashboards and rule chains use them like any other reading.

Result tab of the Site energy this month Period Total model: the predicted total of 271,096.43 for site-energy-001, the chart of actual readings, forecast, and previous period, and a table of the ten site devices

What you can predict

Section titled “What you can predict”

A prediction model is built for one question. Trendz calls this question the intent, and it decides what the model shows, what it writes to ThingsBoard, and which alarms it can raise. The four intents read the same history in four ways, and cover uses such as energy consumption forecasting, fuel and tank levels, battery discharge, filter clogging, occupancy, and production output. Pick one to see how it works.

How it works in Trendz

Section titled “How it works in Trendz”

A model is trained once from the history. After that, every scheduled refresh updates its forecasts and alarms. The forecasting methods and accuracy page explains how the method is chosen, from Prophet and ARIMA to linear regression, and how accuracy is measured.

Blueprint of a prediction model across ThingsBoard, Trendz, and the Python executor: a training task reads history, fits the model in the Python executor, and saves it to the Trendz database; a scheduled refresh starts a refresh task that reads new data, loads the model, optionally retrains, predicts through the Python executor, raises alarms, and saves the forecast back to device telemetry, which dashboards and rule chains use Blueprint of a prediction model across ThingsBoard, Trendz, and the Python executor: a training task reads history, fits the model in the Python executor, and saves it to the Trendz database; a scheduled refresh starts a refresh task that reads new data, loads the model, optionally retrains, predicts through the Python executor, raises alarms, and saves the forecast back to device telemetry, which dashboards and rule chains use

What you get in ThingsBoard:

  • Forecast: saved on each device or asset as telemetry. Show it on a dashboard next to the real readings.
  • Alarms: listed with the other alarms of the device or asset. Send them by email or SMS with a rule chain.

Where to start

Section titled “Where to start”
© 2026 The ThingsBoard Authors