15.4 Train and Score your Recipe

After creating a recipe, you can now train and score your recipe without having to touch code again.

15.4.1 Train a Model based on a Recipe

Login to Adobe Experience Platform: https://experience.adobe.com/platform

DSW

From the left menu, click on Models.

DSW

In this exercise, we’ll use a prebuilt recipe to create a Model for Car Insurance Sales Predictions.

In the top menu, click on Recipes.

DSW

In Recipes, you’ll find multiple recipes. Look for your own recipe in the list, which should be named ldapCarInsurancePropensity.

DSW

Click the ldapCarInsurancePropensity recipe to open it.

DSW

You now need to create your own Model, based on the ldapCarInsurancePropensity recipe.

To do that, click on the Create a Model button.

DSW

To train this model, you need to provide it with an Input Dataset. In our case, the dataset to use is called AEP Demo - Car Insurance Interactions. Select it from the list.

DSW

Click Next to continue.

DSW

In the next step, you need to define a name for your Model. As a naming convention, let’s use: ldap - CarInsurancePropensity Model and replace ldap with your ldap.

Example: for ldap vangeluw, the name become vangeluw - CarInsurancePropensity Model.

DSW

We can also hyper-tune the Model by changing the Model Configuration. To do that, you can for instance change the n_estimators or the max_depth.

DSW

If you want to update the Model’s Configuration parameters, double-click one of the parameters and give it a new value.

Next, click Finish to finish your configuration.

DSW

After a couple of seconds, you’ll be reverted back to the Model’s homepage where you’ll see a Training Run 1 with a status of Pending. The process to finish the training run can take 5+ minutes.

DSW

After 1-2 minutes, your Training Run’s status will change to Running.

DSW

And 1-2 minutes later, the Training Run’s status will change to Complete.

DSW

Once the Training Run has completed, you’ll see an Accuracy Metric that indicates how successful the model is. (Due to the lack of training data, accuracy doesn’t say much in this demo situation).

DSW

Training a model requires more then one run. All Training Runs will be visible on this page and you’ll be able to compare their results, so you can decide which one is the most successful.
DSW

15.4.2 Score a Model based on a Recipe

After training a model, we can use the model to score and as such, have the model calculate Car Insurance Sales Propensity scores which can be activated through targeting.

To start scoring, let’s re-open Training Run 1 by clicking it.

DSW

After opening Training Run 1, you’ll see a full overview of the Training Run, and in the future, more visualization options will be added.

DSW

To score, you have to click the + Score button in the top right corner of your screen.

DSW

In the next step, you again have to select an Input Dataset. Let’s choose the AEP Demo - Car Insurance Interactions dataset.

DSW

After selecting the Input Dataset, click Next.

DSW

In the next step, you need to select a dataset to which Platform will output results. In this case, select the AEP Demo - ML Predictions dataset.

DSW

After selecting the Output Dataset, click Next.

DSW

In the next screen, you can again specify/change some of the Model’s Configuration parameters.

DSW

After updating the Model’s Configuration parameters, click Finish.

DSW

A Scoring Run is now created, and has a status of Running.

DSW

And 1-2 minutes later, the Scoring Run’s status will change to Complete.

DSW

And finally, let’s preview the results. Click on Scoring Run 1

DSW

Next, click the Preview Scoring Results Dataset.

DSW

DSW

Next Step: 15.5 Publish your Recipe as a Scheduled Service

Go Back to Module 15

Go Back to All Modules

On this page

Adobe Summit Banner

A virtual event April 27-28.

Expand your skills and get inspired.

Register for free
Adobe Summit Banner

A virtual event April 27-28.

Expand your skills and get inspired.

Register for free