aws-ml-binary-classification

Lab to illustrate using Amazon Machine Learning to diagnose breast cancer in patients (binary classification)

Getting Started

Deploy the CloudFormation stack in the template in infrastructure/. The template creates a user with the following credentials and minimal required permisisons to complete the Lab:

• Username: student
• Password: password

Instructions

• Get the URL of the CSV data file automatically uploaded to S3
• In the Amazon ML Console, launch the standard setup wizard to:
  • Create a datasource from the CSV file in S3
    • Change the diagnosis column to be of type Binary, and use the id column as a row ID
  • Create a model with default values
  • Create an evaluation with default values
• Inspect the datasource and evaluation statistics
• Perform a real-time prediction by pasting in the following record: 000,13.24,14.06,87.16,563.3,0.09679,0.08229,0.06864,0.04791,0.1685,0.05966,0.2599,0.7686,2.158,22.56,0.008962,0.0136,0.02287,0.01415,0.0188,0.0023,15.11,19.26,99.7,711.2,0.144,0.1773,0.239,0.1288,0.2977,0.07259

Cleaning Up

In the Amazon ML Console, delete the datasource, model, and evaluation. Delete the CloudFormation stack to remove all the remaining template resources.

Acknowledgments

Thank you to the creators of the Wisconsin Diagnostic Breast Cancer dataset and the UC Irvine Machine Learning Repository for hosting the original data for all.