Train and evaluate a model (UI)
In Adobe Experience Platform Data Science Workspace, a machine learning Model is created by incorporating an existing Recipe that is appropriate for the Model's intent. The Model is then trained and evaluated to optimize its operating efficiency and efficacy by fine-tuning its associated Hyperparameters. Recipes are reusable, meaning that multiple Models can be created and tailored to specific purposes with a single Recipe.
This tutorial walks through the steps to create, train, and evaluate a Model.
In order to complete this tutorial, you must have access to Experience Platform. If you do not have access to an IMS Organization in Experience Platform, please speak to your system administrator before proceeding.
This tutorial requires an existing Recipe. If you do not have a Recipe, follow the Import a packaged Recipe in the UI tutorial before continuing.
Create a Model
- In Adobe Experience Platform, click the Models link located in the left navigation column to list all existing Models. Click Create Model near the top right of the page to begin a Model creation process.
- Browse through the list of existing Recipes, find and select the Recipe to be used to create the Model and click Next .
- Select an appropriate input dataset and click Next . This will set the default input training dataset for the Model.
- Provide a name for the Model and review the default Model configurations. Default configurations were applied during Recipe creation, review and modify the configuration values by double-clicking the values. To provide a new set of configurations, click Upload New Config and drag a JSON file containing Model configurations into the browser window. Click Finish to create the Model.Configurations are unique and specific to their intended Recipe, this means that configurations for the Retail Sales Recipe will not work for the Product Recommendations Recipe. See the reference section for a list of Retail Sales Recipe configurations.
Create a training Run
- In Adobe Experience Platform, click the Models link located in the left navigation column to list all existing Models. Find and click on the name of the Model to be trained.
- All existing training runs with their current training statuses are listed. For Models created using the Data Science Workspace user interface, a training run is automatically generated and executed using the default configurations and input training dataset.
- Create a new training run by clicking Train near the top-right of the Model overview page.
- Select the training input dataset for the training run and click Next .
- Default configurations provided during the Model's creation are shown, change and modify these accordingly by double-clicking the values. Click Finish to create and execute the training run.Configurations are unique and specific to their intended Recipe, this means that configurations for the Retail Sales Recipe will not work for the Product Recommendations Recipe. See the reference section for a list of Retail Sales Recipe configurations.
Evaluate the Model
- In Adobe Experience Platform, click the Models link located in the left navigation column to list all existing Models. Find and click on the name of the Model to be evaluated.
- All existing training runs with their current training statuses are listed. With multiple completed training runs, evaluation metrics can be compared across different training runs in the Model evaluation chart, select an evaluation metric using the dropdown list above the graph.The Mean Absolute Percent Error (MAPE) metric expresses accuracy as a percentage of the error. This is used to identify the top performing Experiment. The lower the MAPE, the better.The "Precision" metric describes the percentage of relevant Instances compared with the total retrieved Instances. Precision can be seen as the probability that a randomly selected outcome is correct.Click on a specific training run to view the details of that run. This can be done even before the run has been completed. On the run detail page, you are able to see other evaluation metrics, configuration parameters, and visualizations specific to the training run. You can also download activity logs to see the details of the run. Logs are particularly useful for failed runs to see what went wrong.
- Hyperparameters cannot be trained and a Model must be optimized by testing different combinations of Hyperparameters. Repeat this Model training and evaluation process until you have arrived at an optimized Model.
This tutorial walked you through creating, training, and evaluating a Model in Data Science Workspace. Once you have arrived at an optimized Model, you can use the trained Model to generate insights by following the Score a Model in the UI tutorial.
Retail Sales Recipe configurations
Hyperparameters determine the Model's training behavior, modifying Hyperparameters will affect the Model's accuracy and precision:
Learning rate shrinks the contribution of each tree by learning_rate. There is a trade-off between learning_rate and n_estimators.
The number of boosting stages to perform. Gradient boosting is fairly robust to over-fitting so a large number usually results in better performance.
Maximum depth of the individual regression estimators. The maximum depth limits the number of nodes in the tree. Tune this parameter for best performance; the best value depends on the interaction of the input variables.
Additional parameters determine the Model's technical properties:
List of comma separated input schema attributes.
List of comma separated output schema attributes.
Determines whether input and output features are modifiable
This ID ensures resources you create are namespaced properly and contained within your IMS Organization. Follow the steps here to find your tenant ID.
The input schema used for training a Model.
Column label for evaluation visualizations.
Comma separated list of evaluation metrics to be used for evaluating a Model.
The output schema used for scoring a Model.