8.3 Model Training and Experimentation

So you’ve prepared your data, authored your model and packaged it to test it at scale as a recipe. Now let’s go ahead and train and test the model.

The URL to login to Adobe Experience Platform is: https://experience.adobe.com/platform

8.3.1 - Train a Model based on a Recipe

Log in to Adobe Experience Platform.

After logging in, you’ll land on the homepage of Adobe Experience Platform.

Data Ingestion

Before you continue, you need to select a sandbox. The sandbox to select is named --aepSandboxId--. You can do this by clicking the text Production Prod in the blue line on top of your screen.

Data Ingestion

After selecting the appropriate sandbox, you’ll see the screen change and now you’re in your dedicated sandbox.

Data Ingestion

From the left menu, click on Models.

In this exercise, you’ll use recipe that you created in the previous to make product recommendations.
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 ldapRecommendations.

Click your ldapRecommendations recipe to open it.

DSW

You now need to create your own Model, based on the ldapRecommendations recipe.
To do that, click on the Create Model button.

DSW

To train this model, you need to provide it with an Input Dataset. In our case, you generated data for this Input Dataset in 1 and the input dataset now contains information around product purchase data.
The dataset to use is called Demo System - Event Dataset for Recommendations Model Input (Global v1.1). 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 - Recommendations Model and replace ldap by your ldap.

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

DSW

You can also hyper-tune the Model by changing the Model Configuration. To do that, you can change the number of recommendations and the sampling fraction.

DSW

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

For instance, I’d like to have 3 recommendations with a sampling fraction of 0.8.

DSW

After changing these values, 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 a couple of metrics that indicate the quality of the run:

  • Recall is also known as True Positive Rate and also as Sensitivity: if the real result was Yes, how often did the model predict Yes?

DSW

  • Precision means: When the model predicts Yes, how often is it correct?

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

After training your model, let’s now score it in the next exercise.

Next Step: 8.4 - Scoring and Consumption of Insights

Go Back to Module 8

Go Back to All Modules

On this page