Defaults, negative values

match ← All flat growth benchmarks

96 days around a level below zero.

Input

benchmark/inputs/v1/flat-negative.json: 96 history rows and 24 forecast rows.

Output

Largest difference: 1.1e-11 (< 0.0001% of the input's range).

Every forecast row from both libraries.

Date	Effect Prophet	Python Prophet	Difference
2020-04-06	-35.0017	-35.0017	1.1e-11
2020-04-07	-34.9939	-34.9939	1.1e-11
2020-04-08	-34.9975	-34.9975	1.1e-11
2020-04-09	-34.9977	-34.9977	1.1e-11
2020-04-10	-34.9986	-34.9986	1.1e-11
2020-04-11	-34.9996	-34.9996	1.1e-11
2020-04-12	-35.0005	-35.0005	1.1e-11
2020-04-13	-35.0017	-35.0017	1.1e-11
2020-04-14	-34.9939	-34.9939	1.1e-11
2020-04-15	-34.9975	-34.9975	1.1e-11
2020-04-16	-34.9977	-34.9977	1.1e-11
2020-04-17	-34.9986	-34.9986	1.1e-11
2020-04-18	-34.9996	-34.9996	1.1e-11
2020-04-19	-35.0005	-35.0005	1.1e-11
2020-04-20	-35.0017	-35.0017	1.1e-11
2020-04-21	-34.9939	-34.9939	1.1e-11
2020-04-22	-34.9975	-34.9975	1.1e-11
2020-04-23	-34.9977	-34.9977	1.1e-11
2020-04-24	-34.9986	-34.9986	1.1e-11
2020-04-25	-34.9996	-34.9996	1.1e-11
2020-04-26	-35.0005	-35.0005	1.1e-11
2020-04-27	-35.0017	-35.0017	1.1e-11
2020-04-28	-34.9939	-34.9939	1.1e-11
2020-04-29	-34.9975	-34.9975	1.1e-11

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Convert input rows	13 µs	469 µs	35× faster
Fit	1.83 ms	8.82 ms	4.8× faster
Predict	260 µs	3.11 ms	12× faster
Fit and predict	1.58 ms	12.1 ms	7.6× faster
Convert, fit and predict	1.47 ms	12.7 ms	8.7× faster
First forecast in a new process	176 ms	1.44 s	8.2× faster
Save model to JSON	39 µs	451 µs	11× faster
Load model from JSON	108 µs	2.14 ms	20× faster
Load and predict in a new process	161 ms	1.45 s	9.0× faster

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
  }
}