Skip to content

Repository for credit card fraud detection notebooks and ONNX models

License

Notifications You must be signed in to change notification settings

IBM/ai-on-z-fraud-detection

Repository files navigation

Scope

This repository provides TensorFlow source code for building and training credit card fraud models using an LSTM and a GRU.

Summary

The models included in this repository are multi-layer LSTM or GRU models that analyze time series data to predict whether a credit card transaction is fraudulent.

The models consist of a recurrent neural network (RNN) with 2 layers of long short-term memory (LSTM) or gated recurrent unit (GRU), 200 units in each layer, followed by a dense layer. There is one output, which is Fraud/Non-fraud. A sequence of 7 transactions is used as the input to model.

In a transactional environment such as z/OS CICS, new data is generated from card terminal transactions. The transaction is received by a z/OS application that then must process it. As part of this processing, validation occurs. This can include the use of AI to provide insight on whether the transaction is potentially fraudulent.

The TensorFlow saved models were converted to the ONNX format using the tf2onnx 1.10.0 library and are also included in this repo. The models were converted using the following template:

import tf2onnx
spec = (tf.TensorSpec((7, 16, 220), tf.float32, name="input"),)
output_path = new_model.name + ".onnx"
onnx_model = tf2onnx.convert.from_keras(new_model, spec, output_path=output_path)

Environment:

  • s390x
  • python version 3.6.9
  • pandas version 1.0.1
  • numpy version 1.16.2
  • scikit-learn version 0.22.1
  • sklearn-pandas version 1.8.0
  • TensorFlow version 2.1.0

The dataset used in this repo can be found here:

https://github.com/IBM/TabFormer/tree/main/data/credit_card

License

If you would like to see the detailed LICENSE click here.

About

Repository for credit card fraud detection notebooks and ONNX models

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published