https://github.com/j-sephb-lt-n/unsupervised-fraud-detection

Exploring anomaly detection using unsupervised methods
https://github.com/j-sephb-lt-n/unsupervised-fraud-detection

Last synced: about 23 hours ago
JSON representation

Exploring anomaly detection using unsupervised methods

• Host: GitHub
• URL: https://github.com/j-sephb-lt-n/unsupervised-fraud-detection
• Owner: J-sephB-lt-n
• License: gpl-3.0
• Created: 2024-04-19T07:54:44.000Z (7 months ago)
• Default Branch: main
• Last Pushed: 2024-04-29T08:20:28.000Z (6 months ago)
• Last Synced: 2024-04-29T23:18:35.936Z (6 months ago)
• Language: Python
• Homepage:
• Size: 71.8 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# unsupervised-fraud-detection

Exploring anomaly detection using unsupervised methods

![](./models/evaluation_output/roc_auc.png)

This repo is still massively under construction

Simulate transaction data:

```bash

$ poetry run python -m data.simdata

Simulated dataset written to 'data/input/simdata.csv'

```

Create model training dataset:

```bash

$ poetry run python -m feature_eng.create_train_data

reading input data from 'data/input/simdata.csv'

Finished exporting data to 'feature_eng/output/train_data.csv'

```

Run unsupervised anomaly detection models:

```bash

$ poetry run python -m models.train_predict.dist_to_dst_clust_median

$ poetry run python -m models.train_predict.dist_to_src_clust_median

$ poetry run python -m models.train_predict.local_outlier_factor

$ poetry run python -m models.train_predict.isolation_forest

```

Evaluate models:

```bash

$ poetry run python -m models.evaluate

Exported results to 'models/evaluation_output/'

```

Explain predictions for a specific transaction:

```bash

$ poetry run python -m models.explain_prediction --tid 100420

```