Distress-Model Ensemble

Install and import#

bash

npm install fintech-algorithms

import { distressModelEnsemble } from "fintech-algorithms/fundamental-analysis-and-valuation/integrated-equity-scoring/distress-model-ensemble";

Signature#

distressModelEnsemble(data)

Pools the eligible distress models into one weight-averaged probability and reports how widely those models disagree.

Parameters#

Name	Type	Notes
`data`	`{ models: Array<{ name?: string; eligible: boolean; distress_probability: number; weight: number }> }`	`models` is the candidate list. Only rows whose `eligible` is exactly true contribute; each of those needs a `distress_probability` between 0 and 1 and a positive `weight`, and `name` falls back to `model` when absent. Rows that are not eligible are skipped without validation.

Returns#

{ state: string; method: string; eligible_models: Array<{ name: string; probability: number; weight: number }>; model_count: number; distress_probability: number; weighted_stddev: number; disagreement_range: number; agreement: number; band: string; calibration_boundary: string }

eligible_models lists the contributing rows and model_count their number. distress_probability is the weighted mean, weighted_stddev the weighted standard deviation around it, disagreement_range the highest probability minus the lowest, and agreement one minus that range. band is high-review at 0.66 or above, watch at 0.33 or above, otherwise lower-review.

Errors#

When models is not a nonempty list, or an entry is not an object — throws TypeError
When an eligible model's distress_probability or weight is not a finite number — throws TypeError
When an eligible model's weight is not positive, or its distress_probability is below 0 or above 1 — throws RangeError
When no model is marked eligible — throws RangeError

Complexity: time O(n), space O(n).

Worked example#

executed Captured by running this function on the input its own test provides. Real output of real code — but not asserted against a published figure.

Input#

data

{
  "models": [
    {
      "name": "altman",
      "distress_probability": 0.18,
      "weight": 1,
      "eligible": true
    },
    {
      "name": "ohlson",
      "distress_probability": 0.24,
      "weight": 1,
      "eligible": true
    },
    {
      "name": "zmijewski",
      "distress_probability": 0.3,
      "weight": 0.5,
      "eligible": true
    }
  ]
}

Call#

distressModelEnsemble(data)

Returns#

object with 10 fields: state, method, eligible_models, model_count, distress_probability, weighted_stddev, disagreement_range, agreement, …

{
  "state": "calculated",
  "method": "weighted-distress-probability-ensemble",
  "eligible_models": [
    {
      "name": "altman",
      "probability": 0.18,
      "weight": 1
    },
    {
      "name": "ohlson",
      "probability": 0.24,
      "weight": 1
    },
    {
      "name": "zmijewski",
      "probability": 0.3,
      "weight": 0.5
    }
  ],
  "model_count": 3,
  "distress_probability": 0.22799999999999998,
  "weighted_stddev": 0.04489988864128729,
  "disagreement_range": 0.12,
  "agreement": 0.88,
  "band": "lower-review",
  "calibration_boundary": "weighted aggregation preserves supplied probabilities; it does not recalibrate them"
}

Other exports#

This module also exports calculate, pointInTimeStockScoringInputAssembly, stockScoringPeerCohortResolver, fundamentalMetricDirectionAndPeerNormalization, modelApplicabilityAndVariantRouter, accountingFinancialHealthComposite, earningsQualityComposite, dividendSafetyScore, balanceSheetResilienceScore, crossModelConflictAndDoubleCountingResolver, overallExplainableStockScore, scoreConfidenceMissingDataPenaltyAndAbstention, marketWideStockScreeningAndRanking, stockScoreHistoryMigrationAndChangeAttribution. Every module additionally exports run as an alias of its primary function, and a meta object carrying its catalog id, domain, family, shape and article URL.