Negative level, absmax scaling

match ← All flat growth benchmarks

96 daily rows wobbling around −35.

Input

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

Output

Largest difference: 1.3e-11 (< 0.0001% of the input's range).

Every forecast row from both libraries.

Date	Effect Prophet	Python Prophet	Difference
2020-04-06	-34.9989	-34.9989	1.3e-11
2020-04-07	-34.9989	-34.9989	1.3e-11
2020-04-08	-34.9989	-34.9989	1.3e-11
2020-04-09	-34.9989	-34.9989	1.3e-11
2020-04-10	-34.9989	-34.9989	1.3e-11
2020-04-11	-34.9989	-34.9989	1.3e-11
2020-04-12	-34.9989	-34.9989	1.3e-11
2020-04-13	-34.9989	-34.9989	1.3e-11
2020-04-14	-34.9989	-34.9989	1.3e-11
2020-04-15	-34.9989	-34.9989	1.3e-11
2020-04-16	-34.9989	-34.9989	1.3e-11
2020-04-17	-34.9989	-34.9989	1.3e-11
2020-04-18	-34.9989	-34.9989	1.3e-11
2020-04-19	-34.9989	-34.9989	1.3e-11
2020-04-20	-34.9989	-34.9989	1.3e-11
2020-04-21	-34.9989	-34.9989	1.3e-11
2020-04-22	-34.9989	-34.9989	1.3e-11
2020-04-23	-34.9989	-34.9989	1.3e-11
2020-04-24	-34.9989	-34.9989	1.3e-11
2020-04-25	-34.9989	-34.9989	1.3e-11
2020-04-26	-34.9989	-34.9989	1.3e-11
2020-04-27	-34.9989	-34.9989	1.3e-11
2020-04-28	-34.9989	-34.9989	1.3e-11
2020-04-29	-34.9989	-34.9989	1.3e-11

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Fit	1.91 ms	7.63 ms	4.0× faster
Predict	228 µs	2.69 ms	12× faster
Save model to JSON	22 µs	421 µs	19× faster
Load model from JSON	117 µs	2.08 ms	18× faster
Load and predict in a new process	152 ms	1.41 s	9.2× 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
  }
}