Adaptive API

Regime detection, self-healing forecasts, business constraints, and Forecast DNA.

RegimeDetector

RegimeDetector(nRegimes=2)

Detect statistical regimes using Hidden Markov Models.

Methods

detect(y) → RegimeResult

RegimeResult

Attribute	Type	Description
`.currentRegime`	`str`	Current regime label
`.regimeStats`	`dict`	Per-regime statistics (mean, std, etc.)
`.states`	`np.ndarray`	Regime index per observation
`.labels`	`list[str]`	Regime label per observation
`.transitionMatrix`	`np.ndarray`	K x K transition probability matrix
`.nRegimes`	`int`	Number of detected regimes

RegimeAwareForecaster

RegimeAwareForecaster()

Forecast using the current regime context.

Methods

forecast(y, steps, period=7) → RegimeForecastResult

ForecastDNA

ForecastDNA()

Profile time series characteristics for meta-learning.

from vectrix import ForecastDNA

dna = ForecastDNA()
profile = dna.analyze(values, period=7)

Methods

analyze(y, period=1) → DNAProfile

DNAProfile

Attribute	Type	Description
`.features`	`dict[str, float]`	65+ statistical features
`.fingerprint`	`str`	8-char hex hash
`.difficulty`	`str`	‘easy’, ‘medium’, ‘hard’, ‘very_hard’
`.difficultyScore`	`float`	0–100 score
`.recommendedModels`	`list[str]`	Sorted by fitness
`.category`	`str`	‘trending’, ‘seasonal’, ‘stationary’, etc.
`.summary`	`str`	Natural language summary

!!! warning “Feature values are inside the features dict”

# CORRECT
profile.features['trendStrength']
profile.features['seasonalStrength']

# WRONG — AttributeError
profile.trendStrength
```

## SelfHealingForecast

`SelfHealingForecast(predictions, lower95, upper95, historicalData, period=7, healingMode='adaptive')`

Auto-correct forecasts as actual data arrives.

### Methods

- `observe(actuals)` → `HealingStatus` — Feed actual values as they arrive
- `getUpdatedForecast()` → `Tuple[np.ndarray, np.ndarray, np.ndarray]` — (predictions, lower95, upper95)
- `getReport()` → `HealingReport` — Full healing process report

### HealingReport

| Attribute | Type | Description |
|---|---|---|
| `.overallHealth` | `str` | `excellent`, `good`, `fair`, `poor` |
| `.healthScore` | `float` | 0–100 health score |
| `.totalObserved` | `int` | Number of actual values received |
| `.totalCorrected` | `int` | Number of corrections applied |
| `.improvementPct` | `float` | Error reduction percentage |

## ConstraintAwareForecaster

`ConstraintAwareForecaster()`

Enforce business constraints on predictions.

### Methods

- `apply(predictions, lower95, upper95, constraints, smoothing=True)` → `ConstraintResult`

## Constraint

`Constraint(type, params)`

| Type | Parameters |
|------|-----------|
| `non_negative` | `{}` |
| `range` | `{'min': 0, 'max': 5000}` |
| `capacity` | `{'capacity': 10000}` |
| `yoy_change` | `{'maxPct': 30, 'historicalData': array}` |
| `sum_constraint` | `{'window': 7, 'maxSum': 10000}` |
| `monotone` | `{'direction': 'increasing'}` |
| `custom` | `{'func': callable}` |