The repository is based https://github.com/mikames/data-and-notebooks and I've added the model training and merge prediction with Mike's prediction notebook. I hope that you enjoy the learning.
- Spin up environments with fraud-detector-workshop.yaml, which creates:
- Amazon SageMaker Notebook instance
- IAM roles for Notebook instance and Amazon Fraud Detector data access
- Open Jupyter notebook from SageMaker Notebook instance and start experiencing model training and prediction with Fraud Detector.
If the workshop is being done on a new AWS account or an account without using Amazon SageMaker & Amazon Fraud Detector before, the two month free tier offering should be able to cover the cost. For more detail about free tier / free trial: Amazon SageMaker & Amazon Fraud Detector. Otherwise, you may need to pay fees with on-demand usage and below is the estimation.
- Estimated Cost: $31.5USD = 0.39 * 1 + 0.06 * 1.5 + 0.03 * 1000 + 0.0739 * 1.5 with assumptions:
- The workshop takes approx. 1.5 hours.
- 25MB data file is stored in S3 bucket tempoarily (with which the cost can be ignored).
- Only one model version is created with train data set - project_1_newaccounts_100k.csv, which takes approx. 1hr. And the model hosting time is approx. 1 hours.
- Approx 1,000 times of fraud predictions will be invoked.
- 1 hours usage of t3.medium of Amazon SageMake Notebook instance in Sydney Region ($0.0739 USD per hour)
This repository contains sample notebooks and data, feel free to use as you like.