Google Docs Extension

Actable AI offers you to analyze your data with our provided models directly in your Google Sheets. By following this documentation you will learn

  1. How to install the extension
  2. A case study: use Actable AI to predict a continuous value
  3. Review prediction model

Install the Extension

Before using the plugin, please ensure your browser is only linked with one Google account. There is a known issue about Google add-ons authentication when you have multiple accounts linked in the browser.

Actable AI releases google docs extension on the official marketplace. To install our extension, simply open the sheet you want to analyse, then click on Extensions > Add-ons > Get add-ons

_images/install-plugin-menu.png

Google Marketplace window would pop up on top of your sheet. Search for Actable AI.

_images/install-page.png

Click the install button and grant permission to our plugin. We respect users’ data and would not keep any of your sheets on our servers.

Choose the account you would like to use and allow access.

_images/addon-login.png

Congrats, you have successfully installed the plugin.

_images/install-finish.png

Once the installation is finished, you would see Actable AI plugin listed in your Extensions drop-down menu (Extensions > Actable AI). By clicking the Home button, the side panel would open and our plugin is ready to use.

_images/addon-dropdown-menu.png

Case Study

For example, we have an excel sheet that describes apartment rental prices. A sample of such a sheet would look like the following:

NUMBER_OF_ROOMS SQFT LOCATION DAYS_ON_MARKET INITIAL_PRICE NEIGHBORHOOD RENTAL_PRICE
2 878 great 8 3861 berkeley_hills 3863.323486
2 771 good 8 3305 downtown 3296.687988
1 333 great 6 2284 east_elmwood 2273.651611
1 500 poor 54 1448 northwest 1332.853516
3 1104 poor 16 4750 westbrae  
1 543 poor 18 1871 westbrae  
2 872 good 14 3375 west_welmwood 3373.663818
1 673 great 7 2604 south_side 2583.209473

As you may observe, some rental prices are missing and we’d like to predict them. To achieve this, we can use the regression model provided by Actable AI. Let’s choose all columns except RENTAL_PRICE as the predictors and RENTAL_PRICE as the target. Moreover, we enable the Explain predictions for us to review the model later. Your setting should look like this:

_images/plugin-setup.png

To understand more about the parameters we provide, please refer to this documentation Regression.

Review Result

Please wait patiently, the model would be ready for you shortly. As you may notice, the predicted results would be fulfilled in blue.

_images/data-result.png

Actable AI provides you not only the results but also how confident our model is. Please check the automatically created sheet that would be named as the format <original-sheet-name>_<model>_validation_<date time>

_images/model-validation-sheet-tab.png

Actable AI provides multiple validation metrics for you to evaluate the model just trained, such as Mean Average Error (MAE), Root Mean Square Error (RMSE) and R-squared (R2). Our prediction R2 is approaching 0.999906 in this example.

Except for the validation metrics, Actable AI also tells you which feature contributes the most/least to the target. As you may notice, there is a Feature-Importance table in the validation sheet. In this example, we learned, the INITIAL_PRICE contributes the most to the RENTAL_PRICE. However, to understand more about the causal for the rental price, we would strongly recommend you to use Causal Inference.

_images/validation-metrics.png

The bottom table for this sheet is the validation details. To provide the model confidence, Actable AI would randomly sample a part of the table as the validation set, and run the prediction on this subset of data to calculate the errors. The size of the validation sample is controlled by the slider validation percentage in the control panel.

_images/validation-table.png