Explicit floor with a regressor, absmax scaling

match ← All logistic growth benchmarks

96 daily rows between a floor of about 12 and a ceiling of about 60, with a weekly pattern and a multiplicative promotion regressor.

Input

benchmark/inputs/v1/logistic-explicit-floor.json: 96 history rows and 24 forecast rows.

Output

Largest difference: 1.4e-11 (< 0.0001% of the input's range).

Every forecast row from both libraries.

Date	Effect Prophet	Python Prophet	Difference
2020-04-06	65.8327	65.8327	1.4e-11
2020-04-07	66.2481	66.2481	1.3e-11
2020-04-08	67.2854	67.2854	1.3e-11
2020-04-09	67.5699	67.5699	1.3e-11
2020-04-10	67.9570	67.9570	1.3e-11
2020-04-11	68.0033	68.0033	1.3e-11
2020-04-12	67.8977	67.8977	1.2e-11
2020-04-13	67.9429	67.9429	1.2e-11
2020-04-14	68.6956	68.6956	1.2e-11
2020-04-15	68.9348	68.9348	1.1e-11
2020-04-16	69.5493	69.5493	1.1e-11
2020-04-17	69.8952	69.8952	1.1e-11
2020-04-18	69.9016	69.9016	1.1e-11
2020-04-19	69.7572	69.7572	1.1e-11
2020-04-20	70.1556	70.1556	1.0e-11
2020-04-21	70.0990	70.0990	9.5e-12
2020-04-22	70.6851	70.6851	9.1e-12
2020-04-23	71.2654	71.2654	8.9e-12
2020-04-24	71.5783	71.5783	8.9e-12
2020-04-25	71.5528	71.5528	9.0e-12
2020-04-26	71.7759	71.7759	9.0e-12
2020-04-27	71.3554	71.3554	8.4e-12
2020-04-28	71.6607	71.6607	7.9e-12
2020-04-29	72.2190	72.2190	7.4e-12

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Convert input rows	38 µs	480 µs	13× faster
Fit	21.4 ms	32.1 ms	1.5× faster
Predict	339 µs	3.46 ms	10× faster
Fit and predict	21.5 ms	35.9 ms	1.7× faster
Convert, fit and predict	21.2 ms	36.2 ms	1.7× faster
First forecast in a new process	192 ms	1.44 s	7.5× faster
Save model to JSON	37 µs	481 µs	13× faster
Load model from JSON	103 µs	2.43 ms	24× faster
Load and predict in a new process	153 ms	1.44 s	9.4× faster

Settings

The case declaration both libraries ran.

{
  "kind": "map",
  "comparison": {
    "kind": "equivalent-objective",
    "evidenceId": "logistic-map-v1"
  },
  "configuration": {
    "growth": "logistic",
    "scaling": "absmax",
    "seasonalityMode": "additive",
    "holidaysMode": "additive",
    "changepoints": {
      "mode": "explicit",
      "timestamps": [
        "2020-02-10T00:00:00.000Z"
      ]
    },
    "changepointPriorScale": 0.2,
    "seasonalities": [
      {
        "name": "weekly-custom",
        "periodDays": 7,
        "fourierOrder": 2,
        "priorScale": 10
      }
    ],
    "events": [],
    "regressors": [
      {
        "name": "promotion",
        "priorScale": 10,
        "standardization": "never",
        "mode": "multiplicative"
      }
    ]
  },
  "pythonOptimizer": {
    "algorithm": "Auto",
    "maxIterations": 10000,
    "newtonFallback": true,
    "sigFigs": 12
  }
}