Model Hub

Basic model demonstrating Spark, Elasticsearch, and local CSV data adapters. Supports both v1 (execute) and v2 (train/infer) interfaces for backward compatibility.

Mock model for testing the V2 model interface. Simulates training with 95% accuracy and generates sample inference results. Great reference implementation for building new models.

Isolation Forest anomaly detection using scikit-learn. Identifies statistical outliers using tree-based ensemble methods. Returns risk scores on a 0-100 scale with anomaly classifications.

TensorFlow Dense Autoencoder for reconstruction-error-based anomaly detection. Uses a symmetric encoder-decoder architecture with MSE loss to identify anomalous patterns in numeric data.

PyTorch Autoencoder for reconstruction-error-based anomaly detection. Uses an encoder-decoder architecture with ReLU activations and MSE loss, trained with Adam optimizer.

LSTM Autoencoder for sequential and temporal anomaly detection. Uses Keras LSTM layers with RepeatVector architecture, treating features as time steps for sequence reconstruction.

Graph-based anomaly detection using NetworkX. Constructs graphs from entity relationships and uses PageRank centrality to identify anomalous nodes with high connectivity or influence.