awslabs / fraud-detection-using-machine-learning Goto Github PK

In notebook, the cited link http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.68.6858&rep=rep1&type=pdfor is broken.

I am just trying to run the fraud-detection using machine learning stack and I am encountering this error

UnexpectedStatusException: Error hosting endpoint sagemaker-soln-fd1-rcf: Failed. Reason: Failed to download model data for container "container_1" from URL: "s3://cloud-bucket-test/fraud-classifier/output/sagemaker-soln-fd1-rcf-2021-03-26-18-36-26-322/output/model.tar.gz". Please ensure that there is an object located at the URL and that the role passed to CreateModel has permissions to download the object..

My S3 path where the model.tar.gz is present is given below, my local time is PST

s3://cloud-bucket-test/fraud-classifier/output/sagemaker-soln-fd1-rcf-2021-03-27-00-48-29-687/output/model.tar.gz.

How to change the training container time? Please help

I just want to share the source of the dataset so people can get sense of that dataset mean
https://www.kaggle.com/mlg-ulb/creditcardfraud
Thanks
Ahmed Raafat

Hello!

I'm trying to run through the "sagemaker_fraud_detection" notebook and I'm running into an issue when trying to set the 'content_type' and 'accept' attributes for the different predictors (Random Cut Forest, SMOTE).

Specifically, the commands with the issue:

rcf_predictor.content_type = 'text/csv'
rcf_predictor.serializer = csv_serializer
rcf_predictor.accept = 'application/json'
rcf_predictor.deserializer = json_deserializer

smote_predictor.content_type = 'text/csv'
smote_predictor.serializer = csv_serializer
smote_predictor.deserializer = None

Here is the error that I'm seeing:

AttributeError Traceback (most recent call last)
in
4
5 # Specify input and output formats.
----> 6 smote_predictor.content_type = 'text/csv'
7 smote_predictor.serializer = csv_serializer
8 smote_predictor.deserializer = None

AttributeError: can't set attribute

This issue seems to resolve itself with the random cut forest model but not with the SMOTE model.

Thanks in advance for any insights into this issue, and my apologies if I'm not doing something correctly.

Thanks!

SolutionAssistantStack resource is HTTP 403 on https://s3.us-west-2.amazonaws.com/sagemaker-solutions-build-us-west-2/Fraud-detection-using-machine-learning/build/ecea829bfbfd1c53fd6a004f9537412c.template

Is this solution still available in sagemaker studio under solutions?
I could not find it.

I found the following issues when running this notebook:

1. Metric for evaluation: after running this notebook, I found that the recall was only about 73%, which for a fraud detection use case should be targeted at a higher level such as 90%.

   Suggested resolution: follow the example at https://github.com/awslabs/amazon-sagemaker-examples/blob/master/scientific_details_of_algorithms/linear_learner_class_weights_loss_functions/linear_learner_class_weights_loss_functions.ipynb, including:

   Set 'binary_classifier_model_selection_criteria': 'precision_at_target_recall'
   Split the dataset into validation and test along with train.
   Use class weights etc.
   

2. The data upload cell cannot be run as written. Instead, use the default session bucket or allow the user to provide their own:

     bucket = session.default_bucket()
     prefix = 'fraud-detection-end-to-end-demo/linear-learner'