AI for Sheets
How to install Peltarion AI for Sheets
You can install the Peltarion AI for Sheets add-on directly within an open Google Sheets document:
Click on the menu Add-ons and select Get add-ons
Fill in the Search apps text box with Peltarion - AI for Sheets
Click on the Peltarion - AI for Sheets result, then click on the Install button
How to use Peltarion AI for Sheets
Deploy and enable your model
To deploy a model, go to the Deployment view and click on New deployment. Select the Experiment and Epoch that correspond to the model you want to deploy. If you want, give a custom Name to the deployment, then click on Create.
In the Parameters section, you can change the Name of the input and output features before you enable the deployment for the first time.
The Name specified here is what you will need to use inside Google Sheets.
Anyone with the URL and Token information of a deployment can request predictions from the deployed model.
Don’t share this information or write it inside a document that others can read if they are not allowed to query your deployment.
A deployment can be prevented from accepting requests by disabling it. If you need to get a new token, you can create a new deployment from the same experiment and epoch. Unused deployments can be deleted after they have been disabled.
Get the API information
The Peltarion AI for Sheets add-on needs the
token of the model you want to get predictions from.
This information is available in the Deployment view. Select the deployment you want to use from your Google spreadsheet, and look for the URL and Token values in the API information.
You will also need to know the Name of the input features, as displayed in the Parameters section of the Deployment view.
Request predictions in Google Sheets
In your Google Sheets spreadsheet, click on the Add-ons menu, select Peltarion - AI for Sheets, then Show sidebar. A sidebar will appear on the right side to help you request predictions from your model.
Using the Peltarion AI for Sheets sidebar
Click on Enter credentials to fill in the
Token from the deployment’s API information.
Select the method that corresponds to the type of problem that the model solves.
Regression: when the model returns a single numerical value from a regression problem.
Classification: when the model returns one of several categories from a classification problem.
Binary classification: when the model returns whether or not an example belongs to the positive class of a binary classification problem.
This method adds a Threshold field that lets you specify the value, between 0 and 1, above which an example is considered positive.
Debugging: this method doesn’t process the model predictions, but displays the raw JSON output from the model.
Specify the range of cells in the spreadsheet that has the name of the input features expected by your model. The text in the specified cells must match the Name from the deployment’s Parameters.
Specify the range of cells in the spreadsheet that contains the value of the input features. The data must be organized with one row corresponding to one example to submit to the model, and different columns corresponding to the various features.
Specify the range of cells where the predictions from the model should go. This range should be a column containing as many rows as there are rows in the submitted data range.
Click on Predict to send the specified data to your model and fill the output range of the spreadsheet with the calculated predictions.
You can fill in a cell range easily by first selecting it on your spreadsheet, then clicking the Use selection button in the AI for Sheets toolbar.