Advanced Topics

Advanced features for specialized use cases.

Overview

Topic	Description
Anomaly Injection	Fraud, errors, and process issues
Data Quality Variations	Missing values, typos, duplicates
Graph Export	ML-ready graph formats
Intercompany Processing	Multi-entity transactions
Period Close Engine	Month/quarter/year-end processes
Performance Tuning	Optimization strategies

Feature Matrix

Feature	Use Case	Output
Anomaly Injection	ML training	Labels (CSV)
Data Quality	Testing robustness	Varied data
Graph Export	GNN training	PyG, Neo4j
Intercompany	Consolidation testing	IC pairs
Period Close	Full cycle testing	Closing entries

Enabling Advanced Features

In Configuration

# Anomaly injection
anomaly_injection:
  enabled: true
  total_rate: 0.02
  generate_labels: true

# Data quality variations
data_quality:
  enabled: true
  missing_values:
    rate: 0.01

# Graph export
graph_export:
  enabled: true
  formats:
    - pytorch_geometric
    - neo4j

# Intercompany
intercompany:
  enabled: true

# Period close
period_close:
  enabled: true
  monthly:
    accruals: true
    depreciation: true

Via CLI

Most advanced features are controlled through configuration. Use init to create a base config, then customize:

datasynth-data init --industry manufacturing --complexity medium -o config.yaml
# Edit config.yaml to enable advanced features
datasynth-data generate --config config.yaml --output ./output

Prerequisites

Some advanced features have dependencies:

Feature	Requires
Intercompany	Multiple companies defined
Period Close	`period_months` ≥ 1
Graph Export	Transactions generated
FX	Multiple currencies

Output Files

Advanced features produce additional outputs:

output/
├── labels/                      # Anomaly injection
│   ├── anomaly_labels.csv
│   ├── fraud_labels.csv
│   └── quality_issues.csv
├── graphs/                      # Graph export
│   ├── pytorch_geometric/
│   └── neo4j/
├── consolidation/               # Intercompany
│   ├── eliminations.csv
│   └── ic_pairs.csv
└── period_close/                # Period close
    ├── trial_balances/
    ├── accruals.csv
    └── closing_entries.csv

Performance Impact

Feature	Impact	Mitigation
Anomaly Injection	Low	Post-processing
Data Quality	Low	Post-processing
Graph Export	Medium	Separate phase
Intercompany	Medium	Per-transaction
Period Close	Low	Per-period

SyntheticData Documentation