How aryan9-6-5/neural-volatility-forecaster works
The repository is an implied-volatility-surface forecasting platform. Its main workflow turns options-chain data into standardized surface sequences, trains and evaluates forecasting models, then serves forecasts and drift status through an API and interactive interface. The graph includes documented baseline, statistical-comparison, quality-reporting, and visualization capabilities. Some subsystems are included from the README and source index; their internal behavior and connections are not inferred where excerpts do not show them.
Main entry points: app.py Forecast API · train.py Training & Benchmarking · dataset.py Dataset Builder · processor.py Surface Processor
Components 18
Market Data
- Snapshot Collector
data/collector.py - Surface Processor
data/processor.py - Dataset Builder
data/dataset.py - Quality Reporting
data/historical_quality_report.py - History Buffer
data/rolling_buffer.py
Forecasting
- Forecast Architectures
models/architectures.py - Regularized Loss
models/loss.py - Econometric Baselines
models/baselines.py - Training & Benchmarking
training/train.py - Checkpoint Serialization
models/serialization.py
Serving
- Forecast API
inference/app.py - Prediction Store
inference/prediction_store.py - Drift Monitor
inference/monitor.py - Interactive Interface
inference/static
Evaluation & Visualization
- Significance Tests
training/significance_tests.py - Forecast Visualizations
utils/plotting.py
Other components
- Options User
- Options Market Data
Connections 22
- Options Market Data → Snapshot Collector: provides snapshots
README.md - Snapshot Collector → Surface Processor: collects for processing
- Surface Processor → Dataset Builder: builds surfaces
data/dataset.py - Dataset Builder → Surface Processor: processes snapshots
data/dataset.py - Dataset Builder → Quality Reporting: reports quality
data/historical_quality_report.py - Forecast API → Surface Processor: processes input
inference/app.py - Forecast API → Dataset Builder: builds sequences
inference/app.py - Forecast API → History Buffer: updates history
inference/app.py - Training & Benchmarking → Dataset Builder: loads data
training/train.py - Training & Benchmarking → Forecast Architectures: trains models
training/train.py - Training & Benchmarking → Regularized Loss: optimizes with
training/train.py - Training & Benchmarking → Econometric Baselines: benchmarks against
training/train.py - Training & Benchmarking → Checkpoint Serialization: saves checkpoints
training/train.py - Forecast API → Forecast Architectures: runs forecasts
inference/app.py - Forecast API → Checkpoint Serialization: loads checkpoint
inference/app.py - Forecast API → Prediction Store: stores forecasts
inference/app.py - Forecast API → Drift Monitor: checks drift
inference/app.py - Forecast API → Forecast Visualizations: generates plots
inference/app.py - Forecast API → Training & Benchmarking: triggers retraining
inference/app.py - Options User → Forecast API: requests forecast
README.md - Forecast API → Interactive Interface: serves interface
README.md - Training & Benchmarking → Significance Tests: supplies predictions