keras-io/imbalanced_classification

Model Description

Keras Implementation of Imbalanced classification: credit card fraud detection

This repo contains the trained model of Imbalanced classification: credit card fraud detection.
The full credit goes to: fchollet

Intended uses & limitations

• The trained model is used to detect of a specific transaction is fraudulent or not.

Training dataset

• Credit Card Fraud Detection
• Due to the high imbalance of the target feature (417 frauds or 0.18% of total 284,807 samples), training weight was applied to reduce the False Negatives to the lowest level as possible.

Training procedure

Training hyperparameter

The following hyperparameters were used during training:

• optimizer: ‘Adam’
• learning_rate: 0.01
• loss: ‘binary_crossentropy’
• epochs: 30
• batch_size: 2048
• beta_1: 0.9
• beta_2: 0.999
• epsilon: 1e-07
• training_precision: float32

Training Metrics

Model Plot