Defaults, one month of daily data

match ← All linear growth benchmarks

32 days of history. Weekly seasonality switches on automatically.

Input

benchmark/inputs/v1/linear-small.json: 32 history rows and 16 forecast rows.

Output

Largest difference: 4.4e-8 (< 0.0001% of the input's range).

Every forecast row from both libraries.

Date	Effect Prophet	Python Prophet	Difference
2020-02-02	23.6093	23.6093	3.2e-8
2020-02-03	23.4525	23.4525	4.4e-9
2020-02-04	24.1603	24.1603	3.6e-8
2020-02-05	25.7297	25.7297	2.0e-8
2020-02-06	27.4499	27.4499	8.4e-9
2020-02-07	28.2413	28.2413	1.5e-8
2020-02-08	27.9103	27.9103	6.2e-9
2020-02-09	27.2797	27.2797	2.3e-8
2020-02-10	27.1229	27.1229	1.3e-8
2020-02-11	27.8307	27.8307	4.4e-8
2020-02-12	29.4001	29.4001	2.8e-8
2020-02-13	31.1203	31.1203	1.7e-8
2020-02-14	31.9116	31.9116	2.4e-8
2020-02-15	31.5806	31.5806	2.3e-9
2020-02-16	30.9501	30.9501	1.5e-8
2020-02-17	30.7933	30.7933	2.1e-8

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Convert input rows	5 µs	437 µs	96× faster
Fit	18.2 ms	61.2 ms	3.4× faster
Predict	204 µs	3.15 ms	15× faster
Fit and predict	18.4 ms	64.1 ms	3.5× faster
Convert, fit and predict	18.1 ms	64.9 ms	3.6× faster
First forecast in a new process	191 ms	1.50 s	7.8× faster
Save model to JSON	28 µs	390 µs	14× faster
Load model from JSON	140 µs	2.06 ms	15× faster
Load and predict in a new process	164 ms	1.44 s	8.8× faster

Settings

The case declaration both libraries ran.

{
  "kind": "map",
  "comparison": {
    "kind": "equivalent-objective",
    "evidenceId": "linear-output-first-v4"
  },
  "fitRequest": "growth-only",
  "configuration": {
    "growth": "linear",
    "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
  }
}