Part A: Build, train, and deploy an Amazon Fraud Detector model

In part A, you define your business use case, define your event, build a model, train the model, evaluate model's performance, and deploy the model.

Step 1: Choose your business use case

• In this step, you use the data models explorer to match your business use case with the fraud detection model types supported by Amazon Fraud Detector. Data models explorer is a tool integrated with the Amazon Fraud Detector console that recommends a model type to use for creating and training a fraud detection model for your business use case. Data models explorer also provides insights into the mandatory, recommended, and optional data elements you will require to include in your dataset. The dataset will be used to create and train your fraud detection model.

  For the purpose of this tutorial, your business use case is new account registrations. After you specify your business use case, the data models explorer will recommend a model type for creating a fraud detection model and will also provide you with a list of data elements you will need to create your dataset. Since you have already uploaded a sample dataset containing data from new account registrations, you do not need to create a new dataset.

  1. Open the AWS Management Console and sign in to your account. Navigate to Amazon Fraud Detector.

  2. In the left navigation pane, choose Data models explorer.

  3. In the Data models explorer page, under Business use case, select New account fraud.

  4. Amazon Fraud Detector displays the recommended model type to use to create a fraud detection model for the selected business use case. The model type defines the algorithms, enrichments, and transformations Amazon Fraud Detector will use to train your fraud detection model.

     Make a note of the recommended model type. You will need this later when you create your model.

  5. The Data model insights pane provides insight into the mandatory and recommended data elements required to create and train a fraud detection model.

     Take a look at the sample dataset you downloaded and make sure that it has all the mandatory and some recommended data elements listed in the table.

     Later when you create a model for your specific business use case, you will use the insights provided to create your dataset.

Step 2: Create event type

• In this step, you define the business activity (event) to evaluate for fraud. Defining the event involves setting the variables that are in your dataset, the entity initiating event, and the labels that classify the event. For this tutorial, you define the account registration event.

  1. Open the AWS Management Console and sign in to your account. Navigate to Amazon Fraud Detector.

  2. In the left navigation pane, choose Events.

  3. In the Events type page, choose Create.

  4. Under Event type details, enter sample_registration as the event type name and, optionally, enter a description of the event.

  5. For Entity, choose Create entity.

  6. In the Create entity page, enter sample_customer as the entity type name. Optionally, enter a description of the entity type.

  7. Choose Create entity.

  8. Under Event variables, for Choose how to define this event's variables, choose Select variables from a training dataset.

  9. For IAM role, choose Create IAM role.

  10. In the Create IAM role page, enter the name of the S3 bucket that you uploaded your example data to and choose Create role.

  11. In Data location, enter the path to your example data. This is the S3 URI path that you saved after uploading the example data. The path is similar to this: S3://your-bucket-name/example dataset filename.csv.

  12. Choose Upload.

      Amazon Fraud Detector extracts the headers from your example data file and maps them with a variable type. The mapping is displayed in the console.

  13. Under Labels - optional, for Labels, choose Create new labels.

  14. In Create label page, enter fraud as the name. This label corresponds to the value that represents the fraudulent account registration in the example dataset.

  15. Choose Create label.

  16. Create a second label, then enter legit as the name. This label corresponds to the value that represents the legitimate account registration in the example dataset.

  17. Choose Create event type.

Step 3: Create model

1. On the Models page, choose Add model, and then choose Create model.

2. For Step 1 – Define model details, enter sample_fraud_detection_model as the model name. Optionally, add a description of the model.

3. For Model Type, choose the Online Fraud Insights model.

4. For Event type, choose sample_registration. This is the event type that you created in Step 1.

5. In Historical event data,

   1. In Event data source, choose Event data stored in S3.

   2. For IAM role, select the role that you created in Step 1.

   3. In Training data location, enter the S3 URI path to your example data file.

6. Choose Next.

Step 4: Train model

1. In Model inputs, leave all checkboxes checked. By default, Amazon Fraud Detector uses all variables from your historical event dataset as model inputs.

2. In Label classification, for Fraud labels choose fraud as this label corresponds to the value that represents fraudulent events in the example dataset. For Legitimate labels, choose legit as this label corresponds to the value that represents legitimate events in the example dataset.

3. For the Unlabeled events treatment, keep the default selection Ignore unlabeled events for this example dataset.

4. Choose Next.

5. After reviewing, choose Create and train model. Amazon Fraud Detector creates a model and begins to train a new version of the model.

   In Model versions the Status column indicates the status of model training. Model training that uses the example dataset takes approximately 45 minutes to complete. The status changes to Ready to deploy after model training is complete.

Step 5: Review model performance

An important step in using Amazon Fraud Detector is to assess the accuracy of your model using model scores and performance metrics. After model training is complete, Amazon Fraud Detector validates model performance using the 15% of your data that wasn't used to train the model and generates a model performance score and other performance metrics.

1. To view model's performance,

   1. In the left navigation pane of the Amazon Fraud Detector console, choose Models.

   2. In the Models page, choose the model that you just trained (sample_fraud_detection_model), and then choose 1.0. This is the version Amazon Fraud Detector created of your model.

2. Look at the Model performance overall score and all other metrics that Amazon Fraud Detector generated for this model.

   To learn more about the model performance score and performance metrics on this page, see Model scores and Model performance metrics.

   You can expect all your trained Amazon Fraud Detector models to have real-world fraud detection performance metrics that are similar to the performance metrics that you see for the model in this tutorial.

Step 6: Deploy model

After you reviewed the performance metrics of your trained model and are ready to use it generate fraud predictions, you can deploy the model.

1. In left navigation pane of the Amazon Fraud Detector console, choose Models.

2. In the Models page, choose sample_fraud_detection_model, and then choose the specific model version that you want to deploy. For this tutorial, choose 1.0.

3. On the Model version page, choose Actions and then choose Deploy model version.

4. In the Model versions, the Status shows the status of the deployment. The status changes to Active after the deployment completes. This indicates that the model version is activated and available to generate fraud predictions. Continue with Part B: Generate fraud predictions to complete steps for generating fraud predictions.