How to label time series efficiently – and boost your AI

Data labeling is a critical step in building high-quality AI models. This blog explains how to speed up the labeling process of time series data from sensors and IoT devices.

Sponsored Post.


The increasing digitization of machines and production processes opens up many exciting possibilities, ranging from early fault detection to usage-based pricing. Such applications build on real-time analytics of sensor data for identifying different states of machine operation.

In order to make a machine learning algorithm recognize meaningful operation states from sensor data, the information about the states must be explicitly available for sensor data from the past. For example, training an AI model requires information such as: State A occurred from 3:10 to 3:17 and from 5:23 to 5:35, State B occurred from 7:28 to 8:11. Oftentimes, data about these states does not exist, but must be generated before training the AI. This process is called data labeling. The quality of the labeling has a massive impact on the quality of the AI! Correcting labeling errors should thus be a first optimization step.

In practice, specific knowledge about the domain and the input data is required for labeling sensor data. In an industrial process, for example, only the process experts can typically interpret patterns in machine data. Therefore, the effort of data labeling can become a significant cost factor of industrial data science projects. Most labeling tools focus on image and text, paying little attention to the specific challenges of massive high-dimensional sensor time series.

Visplore is a graphical tool for interactive labeling and exploring massive multi-variate time series data.

Simply start your labeling workflow directly from Python, Matlab, R or various data sources.

Select and label patterns in time series visualizations and other interactive diagrams with a single click. Use pattern search to label all occurrences of selected patterns automatically – also repeatedly on new data. Audit and correct labels together with domain experts. Finally, retrieve labels from Python, Matlab, and R using a single line of code or export them to Excel/CSV.

Learn more about this interactive approach to sensor data labeling and experience yourself how easy you can label massive time series in Visplore Free! Also watch our free 45-minute webinar on Thursday, September 30, 2021.

Learn more and see it in action