Autoencoder implementation using keras and tensorflow. Ran with python 2.7.16
How to start:
- Explore datasets with data-exploration.py
- Preprocess datasets with data-preprocessing.py
- You are able to run ae.py for regular autoencoder or vae.py for a variational autoencoder.py (after training, the model is saved in the models directory)
- Use title of model inside model-analyser.py to process results
References:
- Ellison, D. (2018) Fraud Detection Using Autoencoders in Keras with a TensorFlow Backend. Retrieved from: https://www.datascience.com/blog/fraud-detection-with-tensorflow
- Schreyer, M. Sattarov, T. Borth, Dengel, A. Reimer, B. (2017). Detection of Anomalies in Large-Scale Accounting Data using Deep Autoencoder Networks Retrieved from: https://arxiv.org/pdf/1709.05254.pdf
- Pumsirirat, A. Yan, L. (2018) Credit Card Fraud Detection using Deep Learning based on Auto-Encoderand Restricted Boltzmann Machine. Retrieved from: https://pdfs.semanticscholar.org/01be/7624aa0e0251182593350a984411c2e5128a.pdf
- An, j. Cho, S. Variational Autoencoder based Anomaly Detectionusing Reconstruction Probability. Retrieved from: https://pdfs.semanticscholar.org/0611/46b1d7938d7a8dae70e3531a00fceb3c78e8.pdf