All files / timesfm-core/src/helpers quantile.ts

100% Statements 36/36
100% Branches 8/8
100% Functions 2/2
100% Lines 36/36

Press n or j to go to the next uncovered block, b, p or k for the previous block.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87              1x                         1x 1x 1x 1x 1x                               1x 8x 8x 8x 8x 8x 2x 2x 2x 2x 8x 1x 1x 1x 5x 5x                                 1x 2x 2x 2x 2x 2x 2x 1x 1x 1x 1x 1x 1x 1x  
/**
 * Quantile helpers for working with TimesFM forecast output.
 *
 * Provides ergonomic access to per-series quantile forecasts
 * without manual Float32Array[][] indexing.
 */
 
import { QUANTILE_INDICES, type ForecastOutput } from '../types';
 
/**
 * Confidence level → quantile index mapping.
 *
 * TimesFM 2.5 provides 9 quantiles (q10 through q90).  The widest
 * prediction interval available is Q10–Q90, which provides **≥ 80%**
 * coverage.  Higher confidence levels (0.9, 0.95) map to the same
 * Q10–Q90 bounds because the model does not output finer-grained
 * percentiles.  Users needing tighter intervals should note that
 * 90% and 95% coverage is NOT guaranteed — these are at-least-80%
 * bounds labeled for convenience.
 */
const CI_MAP: Record<number, { lower: number; upper: number }> = {
  0.8: { lower: QUANTILE_INDICES.Q10, upper: QUANTILE_INDICES.Q90 },
  0.9: { lower: QUANTILE_INDICES.Q10, upper: QUANTILE_INDICES.Q90 },
  0.95: { lower: QUANTILE_INDICES.Q10, upper: QUANTILE_INDICES.Q90 },
};
 
/**
 * Extract a specific quantile forecast for a single series.
 *
 * Zero-copy — returns a direct reference to the underlying Float32Array.
 *
 * @param output        Forecast output from `model.forecast()`.
 * @param seriesIndex   Index of the series (0-based).
 * @param quantileIndex Quantile index (use QUANTILE_INDICES constants).
 *
 * @example
 * ```typescript
 * const q10 = getQuantile(output, 0, QUANTILE_INDICES.Q10);
 * ```
 */
export function getQuantile(
  output: ForecastOutput,
  seriesIndex: number,
  quantileIndex: number,
): Float32Array {
  if (seriesIndex < 0 || seriesIndex >= output.pointForecast.length) {
    throw new RangeError(
      `seriesIndex ${seriesIndex} out of range [0, ${output.pointForecast.length})`,
    );
  }
  if (quantileIndex < 0 || quantileIndex >= output.quantileForecast[seriesIndex].length) {
    const max = output.quantileForecast[seriesIndex].length;
    throw new RangeError(`quantileIndex ${quantileIndex} out of range [0, ${max})`);
  }
  return output.quantileForecast[seriesIndex][quantileIndex];
}
 
/**
 * Get a prediction interval for a single series.
 *
 * Returns the lower and upper bounds of the prediction interval
 * at the given confidence level.  80% CI maps to Q10–Q90.
 *
 * @param output        Forecast output from `model.forecast()`.
 * @param seriesIndex   Index of the series (0-based).
 * @param confidence    Confidence level: 0.8, 0.9, or 0.95.
 *
 * @example
 * ```typescript
 * const { lower, upper } = getPredictionInterval(output, 0, 0.8);
 * ```
 */
export function getPredictionInterval(
  output: ForecastOutput,
  seriesIndex: number,
  confidence: 0.8 | 0.9 | 0.95,
): { lower: Float32Array; upper: Float32Array } {
  const mapping = CI_MAP[confidence];
  if (!mapping) {
    throw new RangeError(`Unsupported confidence: ${confidence}. Supported: 0.8, 0.9, 0.95`);
  }
  return {
    lower: getQuantile(output, seriesIndex, mapping.lower),
    upper: getQuantile(output, seriesIndex, mapping.upper),
  };
}