How trongnguyen1010/risk-adjusted-momentum-clustering works
DELTA is a Vietnam-stock research and product system. A researcher configures an ingestion run; source adapters acquire immutable market and raw financial data, which is normalized, reconciled, quality-checked, and converted into point-in-time features. Experiments cluster securities, measure temporal transitions, and optionally run return-space portfolio backtests. A product builder publishes versioned artifacts, and the HTTP server exposes them to the dashboard for ticker, peer, cluster, quality, and provenance exploration. Some builder and orchestration links are inferred from documented roles because their source was not sampled.
Components 22
Data Foundation
- Ingestion Pipeline
src/delta_t1/pipeline.py - Source Adapters
src/delta_t1/ingestion/sources - Immutable Raw Store
src/delta_t1/ingestion/representative_pilot.py - Normalization
src/delta_t1/ingestion/normalization - Reconciliation
src/delta_t1/ingestion/reconciliation - Quality Gate
src/delta_t1/ingestion/quality.py
Research Core
- Feature Builder
src/delta_t1/features - Experiment Runner
src/delta_t1/experiments/runner.py - Clustering Registry
src/delta_t1/clustering/registry.py
Evaluation
- Temporal Analysis
src/delta_t1/evaluation/temporal_metrics.py - Portfolio Targets
src/delta_t1/backtest/portfolio.py - Returns Engine
src/delta_t1/backtest/returns_engine.py - Research Metrics
src/delta_t1/evaluation
Product Layer
- Artifact Builder
src/delta_t1/product/builder.py - Product Bundle
src/delta_t1/product
Serving
- Product Server
src/delta_t1/product/server.py - Product Repository
src/delta_t1/product/repository.py - Web Dashboard
web/app.js
Other components
- Researcher
- Market Sources
- Financial Sources
- Dashboard User
Connections 23
- Researcher → Ingestion Pipeline: configures run
- Market Sources → Source Adapters: supply market data
- Financial Sources → Source Adapters: supply raw reports
- Source Adapters → Immutable Raw Store: store responses
- Ingestion Pipeline → Source Adapters: runs acquisition
- Ingestion Pipeline → Normalization: normalizes tables
- Normalization → Reconciliation: passes candidates
- Reconciliation → Quality Gate: resolves conflicts
- Ingestion Pipeline → Quality Gate: checks coverage
- Quality Gate → Feature Builder: enables features
- Feature Builder → Experiment Runner: provides snapshots
- Experiment Runner → Clustering Registry: selects algorithm
- Experiment Runner → Temporal Analysis: compares clusters
- Experiment Runner → Portfolio Targets: creates targets
- Experiment Runner → Returns Engine: runs backtest
- Experiment Runner → Research Metrics: computes metrics
- Artifact Builder → Product Bundle: publishes artifacts
- Experiment Runner → Artifact Builder: projects results
- Product Server → Product Repository: loads bundle
- Product Repository → Product Bundle: reads artifacts
- Dashboard User → Web Dashboard: chooses ticker
- Web Dashboard → Product Server: requests API
- Product Server → Web Dashboard: returns detail