Weekly pattern, absmax scaling

match ← All logistic growth benchmarks

96 daily rows rising toward a ceiling of about 60, with a custom weekly pattern and one changepoint. Values are divided by their largest distance from the floor.

Input

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

Output

Largest difference: 1.2e-11 (< 0.0001% of the input's range).

Every forecast row from both libraries.

Date	Effect Prophet	Python Prophet	Difference
2020-04-06	64.6830	64.6830	1.2e-11
2020-04-07	65.1412	65.1412	1.2e-11
2020-04-08	65.8511	65.8511	1.1e-11
2020-04-09	66.5504	66.5504	1.1e-11
2020-04-10	66.9796	66.9796	1.1e-11
2020-04-11	67.0695	67.0695	1.0e-11
2020-04-12	67.0029	67.0029	1.0e-11
2020-04-13	67.0800	67.0800	1.0e-11
2020-04-14	67.4829	67.4829	1.0e-11
2020-04-15	68.1389	68.1389	9.5e-12
2020-04-16	68.7859	68.7859	9.1e-12
2020-04-17	69.1642	69.1642	8.8e-12
2020-04-18	69.2049	69.2049	8.8e-12
2020-04-19	69.0904	69.0904	8.8e-12
2020-04-20	69.1213	69.1213	8.7e-12
2020-04-21	69.4793	69.4793	8.3e-12
2020-04-22	70.0919	70.0919	7.9e-12
2020-04-23	70.6969	70.6969	7.5e-12
2020-04-24	71.0346	71.0346	7.3e-12
2020-04-25	71.0359	71.0359	7.2e-12
2020-04-26	70.8835	70.8835	7.3e-12
2020-04-27	70.8777	70.8777	7.2e-12
2020-04-28	71.2002	71.2002	6.9e-12
2020-04-29	71.7786	71.7786	6.5e-12

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Convert input rows	18 µs	486 µs	26× faster
Fit	19.2 ms	28.5 ms	1.5× faster
Predict	229 µs	3.24 ms	14× faster
Fit and predict	18.9 ms	32.6 ms	1.7× faster
Convert, fit and predict	18.6 ms	33.4 ms	1.8× faster
First forecast in a new process	180 ms	1.49 s	8.3× faster
Save model to JSON	29 µs	421 µs	14× faster
Load model from JSON	89 µs	2.27 ms	25× faster
Load and predict in a new process	159 ms	1.46 s	9.2× 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": []
  },
  "pythonOptimizer": {
    "algorithm": "Auto",
    "maxIterations": 10000,
    "newtonFallback": true,
    "sigFigs": 12
  }
}