Anomalies in a history

match ← All flat growth benchmarks

anomalies(history, { seed }) against fitting in Python, predicting the same history and comparing each value with its 99% range. 180 daily rows of a flat level with three added outliers, which both libraries must flag.

Input

benchmark/inputs/v7/flat-anomalies.json: 180 history rows and 0 forecast rows.

Output

Largest difference: 6.6e-12 (< 0.0001% of the input's range).

Every forecast row from both libraries.

Date	Effect Prophet	Python Prophet	Difference
2020-01-01	34.9824	34.9824	5.6e-12
2020-01-02	36.0757	36.0757	4.4e-12
2020-01-03	36.6987	36.6987	3.3e-12
2020-01-04	35.8520	35.8520	3.4e-12
2020-01-05	34.0579	34.0579	5.4e-12
2020-01-06	33.6349	33.6349	6.4e-12
2020-01-07	33.9415	33.9415	6.6e-12
2020-01-08	34.9824	34.9824	5.6e-12
2020-01-09	36.0757	36.0757	4.4e-12
2020-01-10	36.6987	36.6987	3.3e-12
2020-01-11	35.8520	35.8520	3.4e-12
2020-01-12	34.0579	34.0579	5.4e-12
2020-01-13	33.6349	33.6349	6.4e-12
2020-01-14	33.9415	33.9415	6.6e-12
2020-01-15	34.9824	34.9824	5.6e-12
2020-01-16	36.0757	36.0757	4.4e-12
2020-01-17	36.6987	36.6987	3.3e-12
2020-01-18	35.8520	35.8520	3.4e-12
2020-01-19	34.0579	34.0579	5.4e-12
2020-01-20	33.6349	33.6349	6.4e-12
2020-01-21	33.9415	33.9415	6.6e-12
2020-01-22	34.9824	34.9824	5.6e-12
2020-01-23	36.0757	36.0757	4.4e-12
2020-01-24	36.6987	36.6987	3.3e-12
2020-01-25	35.8520	35.8520	3.4e-12
2020-01-26	34.0579	34.0579	5.4e-12
2020-01-27	33.6349	33.6349	6.4e-12
2020-01-28	33.9415	33.9415	6.6e-12
2020-01-29	34.9824	34.9824	5.6e-12
2020-01-30	36.0757	36.0757	4.4e-12
2020-01-31	36.6987	36.6987	3.3e-12
2020-02-01	35.8520	35.8520	3.4e-12
2020-02-02	34.0579	34.0579	5.4e-12
2020-02-03	33.6349	33.6349	6.4e-12
2020-02-04	33.9415	33.9415	6.6e-12
2020-02-05	34.9824	34.9824	5.6e-12
2020-02-06	36.0757	36.0757	4.4e-12
2020-02-07	36.6987	36.6987	3.3e-12
2020-02-08	35.8520	35.8520	3.4e-12
2020-02-09	34.0579	34.0579	5.4e-12
2020-02-10	33.6349	33.6349	6.4e-12
2020-02-11	33.9415	33.9415	6.6e-12
2020-02-12	34.9824	34.9824	5.6e-12
2020-02-13	36.0757	36.0757	4.4e-12
2020-02-14	36.6987	36.6987	3.3e-12
2020-02-15	35.8520	35.8520	3.4e-12
2020-02-16	34.0579	34.0579	5.4e-12
2020-02-17	33.6349	33.6349	6.4e-12
2020-02-18	33.9415	33.9415	6.6e-12
2020-02-19	34.9824	34.9824	5.6e-12
2020-02-20	36.0757	36.0757	4.4e-12
2020-02-21	36.6987	36.6987	3.3e-12
2020-02-22	35.8520	35.8520	3.4e-12
2020-02-23	34.0579	34.0579	5.4e-12
2020-02-24	33.6349	33.6349	6.4e-12
2020-02-25	33.9415	33.9415	6.6e-12
2020-02-26	34.9824	34.9824	5.6e-12
2020-02-27	36.0757	36.0757	4.4e-12
2020-02-28	36.6987	36.6987	3.3e-12
2020-02-29	35.8520	35.8520	3.4e-12
2020-03-01	34.0579	34.0579	5.4e-12
2020-03-02	33.6349	33.6349	6.4e-12
2020-03-03	33.9415	33.9415	6.6e-12
2020-03-04	34.9824	34.9824	5.6e-12
2020-03-05	36.0757	36.0757	4.4e-12
2020-03-06	36.6987	36.6987	3.3e-12
2020-03-07	35.8520	35.8520	3.4e-12
2020-03-08	34.0579	34.0579	5.4e-12
2020-03-09	33.6349	33.6349	6.4e-12
2020-03-10	33.9415	33.9415	6.6e-12
2020-03-11	34.9824	34.9824	5.6e-12
2020-03-12	36.0757	36.0757	4.4e-12
2020-03-13	36.6987	36.6987	3.3e-12
2020-03-14	35.8520	35.8520	3.4e-12
2020-03-15	34.0579	34.0579	5.4e-12
2020-03-16	33.6349	33.6349	6.4e-12
2020-03-17	33.9415	33.9415	6.6e-12
2020-03-18	34.9824	34.9824	5.6e-12
2020-03-19	36.0757	36.0757	4.4e-12
2020-03-20	36.6987	36.6987	3.3e-12
2020-03-21	35.8520	35.8520	3.4e-12
2020-03-22	34.0579	34.0579	5.4e-12
2020-03-23	33.6349	33.6349	6.4e-12
2020-03-24	33.9415	33.9415	6.6e-12
2020-03-25	34.9824	34.9824	5.6e-12
2020-03-26	36.0757	36.0757	4.4e-12
2020-03-27	36.6987	36.6987	3.3e-12
2020-03-28	35.8520	35.8520	3.4e-12
2020-03-29	34.0579	34.0579	5.4e-12
2020-03-30	33.6349	33.6349	6.4e-12
2020-03-31	33.9415	33.9415	6.6e-12
2020-04-01	34.9824	34.9824	5.6e-12
2020-04-02	36.0757	36.0757	4.4e-12
2020-04-03	36.6987	36.6987	3.3e-12
2020-04-04	35.8520	35.8520	3.4e-12
2020-04-05	34.0579	34.0579	5.4e-12
2020-04-06	33.6349	33.6349	6.4e-12
2020-04-07	33.9415	33.9415	6.6e-12
2020-04-08	34.9824	34.9824	5.6e-12
2020-04-09	36.0757	36.0757	4.4e-12
2020-04-10	36.6987	36.6987	3.3e-12
2020-04-11	35.8520	35.8520	3.4e-12
2020-04-12	34.0579	34.0579	5.4e-12
2020-04-13	33.6349	33.6349	6.4e-12
2020-04-14	33.9415	33.9415	6.6e-12
2020-04-15	34.9824	34.9824	5.6e-12
2020-04-16	36.0757	36.0757	4.4e-12
2020-04-17	36.6987	36.6987	3.3e-12
2020-04-18	35.8520	35.8520	3.4e-12
2020-04-19	34.0579	34.0579	5.4e-12
2020-04-20	33.6349	33.6349	6.4e-12
2020-04-21	33.9415	33.9415	6.6e-12
2020-04-22	34.9824	34.9824	5.6e-12
2020-04-23	36.0757	36.0757	4.4e-12
2020-04-24	36.6987	36.6987	3.3e-12
2020-04-25	35.8520	35.8520	3.4e-12
2020-04-26	34.0579	34.0579	5.4e-12
2020-04-27	33.6349	33.6349	6.4e-12
2020-04-28	33.9415	33.9415	6.6e-12
2020-04-29	34.9824	34.9824	5.6e-12
2020-04-30	36.0757	36.0757	4.4e-12
2020-05-01	36.6987	36.6987	3.3e-12
2020-05-02	35.8520	35.8520	3.4e-12
2020-05-03	34.0579	34.0579	5.4e-12
2020-05-04	33.6349	33.6349	6.4e-12
2020-05-05	33.9415	33.9415	6.6e-12
2020-05-06	34.9824	34.9824	5.6e-12
2020-05-07	36.0757	36.0757	4.4e-12
2020-05-08	36.6987	36.6987	3.3e-12
2020-05-09	35.8520	35.8520	3.4e-12
2020-05-10	34.0579	34.0579	5.4e-12
2020-05-11	33.6349	33.6349	6.4e-12
2020-05-12	33.9415	33.9415	6.6e-12
2020-05-13	34.9824	34.9824	5.6e-12
2020-05-14	36.0757	36.0757	4.4e-12
2020-05-15	36.6987	36.6987	3.3e-12
2020-05-16	35.8520	35.8520	3.4e-12
2020-05-17	34.0579	34.0579	5.4e-12
2020-05-18	33.6349	33.6349	6.4e-12
2020-05-19	33.9415	33.9415	6.6e-12
2020-05-20	34.9824	34.9824	5.6e-12
2020-05-21	36.0757	36.0757	4.4e-12
2020-05-22	36.6987	36.6987	3.3e-12
2020-05-23	35.8520	35.8520	3.4e-12
2020-05-24	34.0579	34.0579	5.4e-12
2020-05-25	33.6349	33.6349	6.4e-12
2020-05-26	33.9415	33.9415	6.6e-12
2020-05-27	34.9824	34.9824	5.6e-12
2020-05-28	36.0757	36.0757	4.4e-12
2020-05-29	36.6987	36.6987	3.3e-12
2020-05-30	35.8520	35.8520	3.4e-12
2020-05-31	34.0579	34.0579	5.4e-12
2020-06-01	33.6349	33.6349	6.4e-12
2020-06-02	33.9415	33.9415	6.6e-12
2020-06-03	34.9824	34.9824	5.6e-12
2020-06-04	36.0757	36.0757	4.4e-12
2020-06-05	36.6987	36.6987	3.3e-12
2020-06-06	35.8520	35.8520	3.4e-12
2020-06-07	34.0579	34.0579	5.4e-12
2020-06-08	33.6349	33.6349	6.4e-12
2020-06-09	33.9415	33.9415	6.6e-12
2020-06-10	34.9824	34.9824	5.6e-12
2020-06-11	36.0757	36.0757	4.4e-12
2020-06-12	36.6987	36.6987	3.3e-12
2020-06-13	35.8520	35.8520	3.4e-12
2020-06-14	34.0579	34.0579	5.4e-12
2020-06-15	33.6349	33.6349	6.4e-12
2020-06-16	33.9415	33.9415	6.6e-12
2020-06-17	34.9824	34.9824	5.6e-12
2020-06-18	36.0757	36.0757	4.4e-12
2020-06-19	36.6987	36.6987	3.3e-12
2020-06-20	35.8520	35.8520	3.4e-12
2020-06-21	34.0579	34.0579	5.4e-12
2020-06-22	33.6349	33.6349	6.4e-12
2020-06-23	33.9415	33.9415	6.6e-12
2020-06-24	34.9824	34.9824	5.6e-12
2020-06-25	36.0757	36.0757	4.4e-12
2020-06-26	36.6987	36.6987	3.3e-12
2020-06-27	35.8520	35.8520	3.4e-12
2020-06-28	34.0579	34.0579	5.4e-12

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Fit	1.83 ms	6.85 ms	3.7× faster
Simulate uncertainty	6.39 ms	123 ms	Not compared: different work
Fit and find anomalies	10.4 ms	133 ms	13× faster

Some steps aren't compared, because the two libraries' public calls produce different outputs. Their times show what each call costs, not which library is faster.

• Simulate uncertainty: Python's predict(vectorized=False) also builds the point forecast and every component, while Effect Prophet's predictUncertainty returns only the ranges.

Settings

The case declaration both libraries ran.

{
  "kind": "map",
  "comparison": {
    "kind": "equivalent-objective",
    "evidenceId": "flat-growth-reduced-v1"
  },
  "fitRequest": "growth-only",
  "configuration": {
    "growth": "flat",
    "scaling": "absmax",
    "seasonalityMode": "additive",
    "holidaysMode": "additive",
    "changepoints": {
      "mode": "auto",
      "count": 25,
      "range": 0.8
    },
    "changepointPriorScale": 0.05,
    "seasonalities": [
      {
        "name": "weekly",
        "periodDays": 7,
        "fourierOrder": 3,
        "priorScale": 10
      }
    ],
    "events": [],
    "regressors": []
  },
  "pythonOptimizer": {
    "algorithm": "Auto",
    "maxIterations": 10000,
    "newtonFallback": true
  },
  "uncertainty": {
    "seed": 19,
    "samples": 1000,
    "intervalWidth": 0.99,
    "output": "intervals",
    "algorithm": "scalar-continuous-time",
    "evidence": "docs/validation/uncertainty.md#evidence-classes",
    "samplerMemoryLimitBytes": 67108864
  },
  "anomalies": {
    "expected": [
      "2020-01-31T00:00:00.000Z",
      "2020-04-05T00:00:00.000Z",
      "2020-05-30T00:00:00.000Z"
    ]
  }
}