How nandita005/decodex_final works
This is a regulatory-surveillance analysis pipeline: raw exchange-style orders, trades, and suspicious lists become cleaned scrip/date batches, client networks and microstructure signals, then ranked and explainable analyst-review artifacts. The main runner owns the documented stage flow; sampled code confirms key calls, while README statements support several unsampled subsystems. The graph preserves those documented capabilities; links without direct sampled call evidence are omitted or marked unverified.
Main entry points: run_pipeline.py Pipeline runner · signal_fusion.py Signal fusion · outputs Reports and graphs · load_data.py Load and clean
Components 22
Data preparation
- Load and clean
src/preprocessing/load_data.py - Trade cleaning
src/preprocessing/clean_trades.py - Order cleaning
src/preprocessing/clean_orders.py - Scrip/date batches
src/preprocessing/build_scrip_date_batches.py
Detection signals
- Trade graph
src/network_analysis/build_trade_graph.py - Cycle detection
src/network_analysis/cycle_detection.py - Centrality and reciprocity
src/network_analysis/centrality_metrics.py - Network analysis
src/network_analysis - Microstructure features
src/microstructure/price_volume.py
Risk assessment
- Signal fusion
src/reasoning_engine/signal_fusion.py - Risk configuration
src/config.py - Client and member risk
src/risk_scoring/client_risk.py - Mechanism labels
src/classification/mechanism_classifier.py
Evidence and review
- Evidence records
src/explainability/evidence_generator.py - Evidence integration
src/s3/integrate_evidence.py - Composite scoring
src/s3/compute_composite_risk.py - Escalation selection
src/s3/select_escalation.py - Review explanations
src/s3/generate_explanations.py
Other components
- Analyst
- Raw market files
data/raw - Pipeline runner
run_pipeline.py - Reports and graphs
outputs
Connections 31
- Raw market files → Load and clean: loads files
README.md - Pipeline runner → Load and clean: loads data
run_pipeline.py - Load and clean → Trade cleaning: cleans trades
src/preprocessing/load_data.py - Load and clean → Order cleaning: cleans orders
src/preprocessing/load_data.py - Pipeline runner → Scrip/date batches: builds batches
run_pipeline.py - Pipeline runner → Trade graph: builds graph
run_pipeline.py - Trade graph → Trade cleaning: normalizes IDs
src/network_analysis/build_trade_graph.py - Pipeline runner → Cycle detection: detects cycles
run_pipeline.py - Pipeline runner → Centrality and reciprocity: computes metrics
run_pipeline.py - Pipeline runner → Microstructure features: computes features
run_pipeline.py - Pipeline runner → Signal fusion: fuses signals
run_pipeline.py - Signal fusion → Centrality and reciprocity: uses metrics
README.md - Signal fusion → Microstructure features: uses features
README.md - Signal fusion → Cycle detection: uses cycle nodes
README.md - Pipeline runner → Client and member risk: scores clients
run_pipeline.py - Client and member risk → Signal fusion: reads components
src/risk_scoring/client_risk.py - Client and member risk → Risk configuration: validates weights
src/risk_scoring/client_risk.py - Pipeline runner → Mechanism labels: classifies clients
run_pipeline.py - Pipeline runner → Evidence records: generates evidence
run_pipeline.py - Evidence records → Signal fusion: uses signal names
src/explainability/evidence_generator.py - Pipeline runner → Evidence integration: runs stage three
run_pipeline.py - Pipeline runner → Composite scoring: computes final risk
run_pipeline.py - Pipeline runner → Escalation selection: selects escalation
run_pipeline.py - Pipeline runner → Review explanations: writes explanations
run_pipeline.py - Evidence integration → Reports and graphs: writes combined records
src/s3/integrate_evidence.py - Composite scoring → Reports and graphs: writes scores
src/s3/compute_composite_risk.py - Review explanations → Reports and graphs: writes report
src/s3/generate_explanations.py - Pipeline runner → Reports and graphs: writes artifacts
README.md - Trade graph → Reports and graphs: renders graphs
README.md - Network analysis → Analyst: supports review
- Reports and graphs → Analyst: provides review artifacts
README.md