Machine Learning Hub and Notebooks

What is Machine Learning Hub?

TrendMiner’s vision to democratize analytics goes beyond empowering domain experts with self-service tools to also include closer collaboration between different experts when it comes to solving problems. Some of the most complex problems require the introduction of data scientists who bring specialized techniques to the table that allow companies to squeeze the deepest insights out of the available data. Think about advanced statistics and machine learning models.

After a very successful trial program with notebooks, we are proud to announce the full launch of a new industrial Machine Learning Hub (MLHub) for timeseries data, extending analytic and machine learning capabilities of the production client of TrendMiner. With MLHub, data scientists can access TrendMiner data (both raw as well as pre-processed and contextualized data in TrendMiner views) and validate hypotheses or create/train/deploy machine learning models in the new Notebook environment, which other users can apply through machine learning model tags and visualize in DashHub.

How to use the MLHub homepage

Important note: MLHub can only be accessed after access management is set up, for which a separate licence is necessary. For details concerning access management read here. Contact us at TrendMiner if you require further information concerning licensing.


In the top left corner of the TrendMiner environment you find the Hub selection button. Here you can select to which hub you want to navigate.

Click on the "MLHub" button.

In the top right corner, you find the rest of the top bar. This remains the same in all TrendMiner hubs.

Just beneath the top bar, the notebook action buttons are present, to create, load and import notebooks.

On the left side of your screen estate, you will find notebook snippets. These are predefined code blocks to use in your notebooks. See “code snippets” below.

What is Notebook?

TrendMiner's Notebook functionality is a platform that enables users to create and work with advanced tooling beyond the robust built-in capabilities, within the TrendMiner environment.

Why use Notebooks?

With embedded notebooks, you will be able to:

  • Load data from a TrendHub view that has been prepared using the typical built-in TrendMiner capabilities (select set of interesting tags, select timeframes of interest eg. via searches, …)
  • Visualize and analyse your data in different ways not possible within TrendMiner.
  • Carry out automation of analytics via scripting (e.g. repeat analysis over a large range of assets).
  • Create (predictive) tags using custom models (e.g. neural nets or clustering) supported by the typical notebook libraries.

You can make use of the more advanced visualization options which come built-in with the notebook.  The embedded Notebook comes with its very own Notebook tile, so that you can also embed your work in a DashHub dashboard and make it available to your entire organization.

How to use Notebooks

Note: Interpreter - The default interpreter of the notebooks is Python.

On the Homepage of MLHub, you can either create a new notebook, load or import an existing note. A new notebook will be always empty.

Create a notebook

  1. Enter the Home page of MLHub
  2. Click on the "New" button or on the “Create notebook” indication. A panel appears from the right of the screen.ML2.png
  3. Populate the open fields and select a work organizer folder to store your new note.ML3.png
  4. Click on the "Create" button.
  5. Select the kernel configuration you would like to launch.ML4.png

We offer kernels including preinstalled visualization packages. We recommend you launch with the “Python with Visualization Libraries on Kubernetes” kernel.

  • Confirm your selection by clicking “select”.

A new notebook will open onto which you can write your python code. You may like to introduce our boiler-plate code which loads highly recommended packages, as you will need it to read in the TrendMiner content. You can add this code by clicking the snippet button  “initialization script” from the snippet menu on the left.

Loading a notebook

  1. Enter the Home page of MLHub.
  2. Click on the “Load" button. A panel showing the work organiser appears from the right of the screen.
  3. Select the notebook you want to open.
  4. Click on ‘load’ button in bottom right.

Importing a notebook

  1. Enter the Home page of MLHub.
  2. Click on the "Import" button. A panel appears from the right of the screen.
  3. Choose to upload a notebook with your local file browser system or choose to use the work organizer. Only Jupyter notebooks (file extension .ipynb) are supported.
  4. Select "import".

Load TrendMiner data as a dataframe

Note: As of version 2022.R2, only "TrendHub Views" and “ContextHub Views” are available as TrendMiner content. More content will be introduced in later versions.

You can load TrendHub or ContextHub views, for example, of good and/or abnormal operation periods and then compare them using advanced analytics, add them to your TRENDMINER CONTENT list.

  1. Click on the blue “+" button. A side panel will appear from the right.
  2. Select the saved view item(s) you wish to add as a dataset.
  3. Click "Load".ML6.png

Items in the content menu list can be either opened or deleted. To delete click on the x in the menu located on the right of the item to be deleted.

  • Click on the new content to open. This will add the relevant python-code into a new cell at the end of the Notebook.ML7.png

TrendHub Views

More information can be obtained by running the following command in the notebook:


load_view( ): Loads the time series data of a saved TrendHub view into a list of Pandas DataFrames. 

  • One DataFrame is returned per layer in the view.
  • Each DataFrame can have a different set of tags available.
  • The optional parameters [layer_ids] allows you to only load a specified list of layers (identified by the layer ids, as provided by the view_info function).

view_info( ): Collect information about a view based on its ID. This info can be used to fetch data from a view: it lists all the layers that are included in the view. When fetching data from the view you can select layers to be included in the data.


Currently snippets are applicable in python kernels to deploy machine learning models. More information can be found here.


Publishing Notebook Output into a Pipeline

If you have a notebook output cell with an interesting result or visualization you want to share, you can publish it into a notebook pipeline object.


  1. Click the ‘Publish’ button in the top right corner.


  1. Select ALL cells that are relevant for the output you want to publish, including ‘import’ statements and variable declarations. Make sure the last selected cell produces valid output. You can select a cell by clicking anywhere on it, or use the checkboxes in the top right corners.




  1. Click ‘Publish’ again. Here you can give the pipeline a name and specify a storage location in the Work Organizer. This will create a ‘pipeline’ object, which is a sort of mini notebook containing only the selected cells. Pipelines have no relation to their origin notebooks, they are a separate entity.If you update the notebook, you will have to create a new pipeline object as well.


  1. Click the ‘Publish’ button in the bottom right.

This will create a pipeline object containing the code of the cells you have selected.

DashHub: Notebook Output Tiles

DashHub allows any output of a Notebook Paragraph to be shown in a dashboard Notebook Tile.

Creating a Notebook Tile

  1. Goto DashHub and create a new dashboard or open a previously created dashboard.
  2. Click on the "Actions" button. A dropdown menu will appear.
  3. Click on "Add new tile". A side panel will appear from the right.
  4. Click on the "Notebook output" option.
  5. Provide a title for the dashboard tile.
  6. Select the pipeline object from Work Organizer.
  7. Adjust the refresh setting if necessary.
  8. Click "Add new tile".


Note: When sharing a DashHub tile, the underlying views and notebook pipeline objects need to be shared as well.

Remarks and Known Issues

Sharing a notebook will result in a symbolic link to the original notebook. This means that the creator remains the owner of the master data of the notebook. Changes that are made to cells are saved on a FIFO principle and will only be updated upon screen refresh.

The published pipeline objects output objects are not linked to the original notebooks anymore. Once published, changes on the notebook will not be automatically impact the existing notebook pipeline object. This allows users to fine-tune notebook cell outputs, without impacting other user/viewers.

The notebooks and visualizations of notebook outputs are run, relative to your permissions and privileges. That means that some data sources might be blocked, and you only get to process data based on your role.


Enjoying TrendMiner? Share your experience to help the community.
Rate TrendMiner