Values far from zero, absmax scaling

match ← All linear growth benchmarks

96 daily rows offset well above zero, with weekly seasonality and one changepoint.

Input

benchmark/inputs/v1/linear-offset-scaling.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	110.0819	110.0819	1.2e-11
2020-04-07	110.2547	110.2547	1.2e-11
2020-04-08	110.5696	110.5696	1.2e-11
2020-04-09	110.8826	110.8826	1.2e-11
2020-04-10	111.0514	111.0514	1.2e-11
2020-04-11	111.0356	111.0356	1.2e-11
2020-04-12	110.9363	110.9363	1.2e-11
2020-04-13	110.9232	110.9232	1.2e-11
2020-04-14	111.0960	111.0960	1.2e-11
2020-04-15	111.4109	111.4109	1.2e-11
2020-04-16	111.7239	111.7239	1.2e-11
2020-04-17	111.8927	111.8927	1.2e-11
2020-04-18	111.8768	111.8768	1.2e-11
2020-04-19	111.7776	111.7776	1.2e-11
2020-04-20	111.7645	111.7645	1.2e-11
2020-04-21	111.9372	111.9372	1.2e-11
2020-04-22	112.2522	112.2522	1.2e-11
2020-04-23	112.5652	112.5652	1.2e-11
2020-04-24	112.7340	112.7340	1.2e-11
2020-04-25	112.7181	112.7181	1.2e-11
2020-04-26	112.6189	112.6189	1.2e-11
2020-04-27	112.6058	112.6058	1.2e-11
2020-04-28	112.7785	112.7785	1.2e-11
2020-04-29	113.0935	113.0935	1.2e-11

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Convert input rows	15 µs	459 µs	30× faster
Fit	5.55 ms	23.3 ms	4.2× faster
Predict	177 µs	3.02 ms	17× faster
Fit and predict	5.44 ms	26.1 ms	4.8× faster
Convert, fit and predict	5.83 ms	26.8 ms	4.6× faster
First forecast in a new process	170 ms	1.45 s	8.5× faster
Save model to JSON	21 µs	419 µs	20× faster
Load model from JSON	106 µs	2.13 ms	20× faster
Load and predict in a new process	157 ms	1.44 s	9.2× faster

Settings

The case declaration both libraries ran.

{
  "kind": "map",
  "comparison": {
    "kind": "equivalent-objective",
    "evidenceId": "linear-output-first-v4"
  },
  "configuration": {
    "growth": "linear",
    "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": []
  },
  "effectOptimizer": {
    "algorithm": "auto",
    "maxIterations": 10000,
    "fallback": "newton",
    "lbfgs": {
      "historySize": 5,
      "initAlpha": 0.001,
      "tolObj": 1e-12,
      "tolRelObj": 10000,
      "tolGrad": 1e-8,
      "tolRelGrad": 10000000,
      "tolParam": 1e-8
    }
  },
  "pythonOptimizer": {
    "algorithm": "Auto",
    "maxIterations": 10000,
    "newtonFallback": true,
    "sigFigs": 12
  }
}