Partition numeric input data into a training, test, and validation set. Normalize the data into range [0,1]. Build a Keras autoencoder to reconstruct the input data without anomalies (fraudulent transactions). Apply the Keras model to the test set with anomalies. Detect anomalies in the test set as exceptional high reconstruction errors. Find the optimal threshold for the reconstruction error and evaluate the model performance by producing the accuracy statistics of the model.
Workflow
Keras Autoencoder for Fraud Detection Training
Used extensions & nodes
Created with KNIME Analytics Platform version 4.5.0
Legal
By using or downloading the workflow, you agree to our terms and conditions.