Business Intelligence

Forecasting is only the first step. Real-world decision-making requires anomaly detection to clean your data, what-if scenarios for planning, backtesting to validate your approach, and business-specific accuracy metrics that go beyond MAPE. Vectrix’s BI module provides all four.

Anomaly Detection

Before forecasting, identify unusual observations that could distort model training

from vectrix.business import AnomalyDetector

detector = AnomalyDetector()
result = detector.detect(data, method="auto")

print(f"Method: {result.method}")
print(f"Anomalies: {result.nAnomalies}")
print(f"Ratio: {result.anomalyRatio:.1%}")
print(f"Indices: {result.indices}")

Methods: auto, zscore, iqr, rolling

What-If Analysis

Explore hypothetical scenarios against your baseline forecast — essential for budget planning, risk assessment, and stakeholder presentations

from vectrix.business import WhatIfAnalyzer

analyzer = WhatIfAnalyzer()
results = analyzer.analyze(base_predictions, historical_data, [
    {"name": "Optimistic", "trend_change": 0.1},
    {"name": "Pessimistic", "trend_change": -0.15},
    {"name": "Shock", "shock_at": 10, "shock_magnitude": -0.3, "shock_duration": 5},
    {"name": "Level Shift", "level_shift": 0.05},
])

for sr in results:
    print(f"{sr.name}: mean={sr.predictions.mean():.2f}, impact={sr.impact:+.1f}%")

Backtesting

How do you know your forecasting approach works? Walk-forward validation simulates historical performance by repeatedly training and predicting

from vectrix.business import Backtester

from vectrix.engine.ets import AutoETS

bt = Backtester(nFolds=5, horizon=14, strategy='expanding')
result = bt.run(data, lambda: AutoETS())

print(f"Avg MAPE: {result.avgMAPE:.2f}%")
print(f"Avg RMSE: {result.avgRMSE:.2f}")
print(f"Best Fold: #{result.bestFold}")
print(f"Worst Fold: #{result.worstFold}")

for f in result.folds:
    print(f"  Fold {f.fold}: MAPE={f.mape:.2f}%")

modelFactory must be a callable that returns a model instance with .fit(y) and .predict(steps) methods.

Strategies: expanding, sliding

Business Metrics

MAPE and RMSE tell you about statistical accuracy, but businesses need different answers: systematic bias detection, volume-weighted error, and naive baseline comparison:

from vectrix.business import BusinessMetrics

metrics = BusinessMetrics()
result = metrics.calculate(actual, predicted)

print(f"Bias: {result['bias']:+.2f}")
print(f"WAPE: {result['wape']:.2f}%")
print(f"MASE: {result['mase']:.2f}")
print(f"Accuracy: {result['forecastAccuracy']:.1f}%")

Interactive tutorial: marimo run docs/tutorials/en/06_business.py