Predictive Ledger

Evaluation configuration

Prediction target

Field whose value Aito predicts on the held-out test set.

Test set size100

Capped at 300 to keep latency bounded. Larger = more stable accuracy estimate, slower call.

Input features

These map to { $get: "fieldname" } bindings in the where clause.

Model performanceFrom _evaluate aggregate

Accuracy

—

Mean rank

—

lower = better

Geom mean p

—

calibrated conf.

Gain

—

vs baseline

Test samples

—

Train samples

—

Correct

—

Evaluation casesFirst 0 of 0 · green=correct, red=wrong

No cases returned. Pick at least one input field.

The Predictive DB

_evaluateOperator

LivePredictions

0Records

IndexedModel

_evaluate (with cases)

Per-domain evaluation using Aito's _evaluate with select: [..., "cases"]. Pick a domain, a target field to predict, and the input features to condition on. KPIs come from the aggregate response; the cases table shows individual predictions vs. ground truth.

Example query

{
  "testSource": {
    "from": "invoices",
    "where": {
      "customer_id": "CUST-0000"
    },
    "limit": 100
  },
  "evaluate": {
    "from": "invoices",
    "where": {
      "customer_id": "CUST-0000",
      "vendor": {
        "$get": "vendor"
      },
      "amount": {
        "$get": "amount"
      }
    },
    "predict": "gl_code"
  },
  "select": [
    "accuracy",
    "meanRank",
    "testSamples",
    "cases"
  ]
}

Verify yourself

Every claim on this page traces back to an Aito query. No separate model file — Aito predicts directly from its index.

View live schema (JSON) Query API reference

Note: GL codes (4100, 5300, 6200…) are illustrative — real Finnish Liikekirjuri uses different numbering. Chosen for demo readability.

Learn more

API reference: _evaluate aito-demo evaluate.js

Start free trial →