Perfectly constant, absmax scaling

match ← All flat growth benchmarks

96 identical values. Both libraries take Prophet's constant-series shortcut.

Input

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

Output

Largest difference: 0.

Every forecast row from both libraries.

Date	Effect Prophet	Python Prophet	Difference
2020-04-06	35.0000	35.0000	0
2020-04-07	35.0000	35.0000	0
2020-04-08	35.0000	35.0000	0
2020-04-09	35.0000	35.0000	0
2020-04-10	35.0000	35.0000	0
2020-04-11	35.0000	35.0000	0
2020-04-12	35.0000	35.0000	0
2020-04-13	35.0000	35.0000	0
2020-04-14	35.0000	35.0000	0
2020-04-15	35.0000	35.0000	0
2020-04-16	35.0000	35.0000	0
2020-04-17	35.0000	35.0000	0
2020-04-18	35.0000	35.0000	0
2020-04-19	35.0000	35.0000	0
2020-04-20	35.0000	35.0000	0
2020-04-21	35.0000	35.0000	0
2020-04-22	35.0000	35.0000	0
2020-04-23	35.0000	35.0000	0
2020-04-24	35.0000	35.0000	0
2020-04-25	35.0000	35.0000	0
2020-04-26	35.0000	35.0000	0
2020-04-27	35.0000	35.0000	0
2020-04-28	35.0000	35.0000	0
2020-04-29	35.0000	35.0000	0

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Fit	1.16 ms	3.10 ms	2.7× faster
Predict	217 µs	2.64 ms	12× faster
Save model to JSON	17 µs	400 µs	24× faster
Load model from JSON	103 µs	2.04 ms	20× faster
Load and predict in a new process	152 ms	1.45 s	9.5× faster

Settings

The case declaration both libraries ran.

{
  "kind": "map",
  "comparison": {
    "kind": "equivalent-objective",
    "evidenceId": "flat-growth-reduced-v1"
  },
  "configuration": {
    "growth": "flat",
    "scaling": "absmax",
    "seasonalityMode": "additive",
    "holidaysMode": "additive",
    "changepoints": {
      "mode": "auto",
      "count": 25,
      "range": 0.8
    },
    "changepointPriorScale": 0.05,
    "seasonalities": [],
    "events": [],
    "regressors": []
  },
  "pythonOptimizer": {
    "algorithm": "Auto",
    "maxIterations": 10000,
    "newtonFallback": true,
    "sigFigs": 12
  }
}