Module benchmarks 

Benchmark suite definitions for ML evaluation.

Provides standardized benchmark datasets for:

• Anomaly detection (AnomalyBench-1K)
• Fraud detection (FraudDetect-10K)
• Data quality detection (DataQuality-100K)
• Entity matching (EntityMatch-5K)

Each benchmark defines:

• Dataset size and composition
• Ground truth labels
• Evaluation metrics
• Expected baseline performance

Structs

BaselineResult

Expected baseline result for a benchmark.

BenchmarkBuilder

Builder for creating benchmark suites.

BenchmarkSuite

A benchmark suite definition.

CostMatrix

Cost matrix for cost-sensitive evaluation.

DatasetSpec

Dataset specification.

EvaluationSpec

Evaluation specification.

FeatureSet

Feature set for the benchmark.

LeaderboardEntry

Leaderboard entry for benchmark results.

SplitRatios

Train/validation/test split ratios.

Enums

BaselineModelType

Types of baseline models.

BenchmarkTaskType

Types of benchmark tasks.

MetricType

Types of evaluation metrics.

Functions

all_benchmarks

Get all available benchmark suites.

anomaly_bench_1k

AnomalyBench-1K: 1000 transactions with known anomalies.

data_quality_100k

DataQuality-100K: 100K records for data quality detection.

entity_match_5k

EntityMatch-5K: 5K records for entity matching.

fraud_detect_10k

FraudDetect-10K: 10K transactions for fraud detection.

get_benchmark

Get a benchmark by ID.

graph_fraud_10k

GraphFraud-10K: 10K transactions with network structure.