Mixed effects, both libraries on default settings

match ← All flat growth benchmarks

The mixed-effects rows with absmax scaling, where both libraries pick their own optimizer. The other mixed-effects cases pin both to L-BFGS.

Input

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

Output

Largest difference: 0.021 (0.073% of the input's range).

Every forecast row from both libraries.

Date	Effect Prophet	Python Prophet	Difference
2020-04-06	21.3445	21.3492	0.0047
2020-04-07	24.0577	24.0640	0.0064
2020-04-08	36.1988	36.2142	0.015
2020-04-09	45.9285	45.9466	0.018
2020-04-10	34.9902	35.0026	0.012
2020-04-11	41.0592	41.0756	0.016
2020-04-12	28.9157	28.9241	0.0085
2020-04-13	21.3445	21.3492	0.0047
2020-04-14	25.2552	25.2655	0.010
2020-04-15	34.9902	35.0026	0.012
2020-04-16	45.9285	45.9466	0.018
2020-04-17	48.6246	48.6457	0.021
2020-04-18	41.0592	41.0756	0.016
2020-04-19	28.9157	28.9241	0.0085
2020-04-20	36.1877	36.2040	0.016
2020-04-21	24.0577	24.0640	0.0064
2020-04-22	35.0012	35.0127	0.011
2020-04-23	45.9285	45.9466	0.018
2020-04-24	48.6246	48.6457	0.021
2020-04-25	34.9902	35.0026	0.012
2020-04-26	30.1132	30.1256	0.012
2020-04-27	21.3445	21.3492	0.0047
2020-04-28	24.0577	24.0640	0.0064
2020-04-29	35.0012	35.0127	0.011

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Convert input rows	52 µs	507 µs	9.7× faster
Fit	3.61 ms	43.6 ms	12× faster
Predict	287 µs	4.73 ms	16× faster
Fit and predict	3.11 ms	48.8 ms	16× faster
Convert, fit and predict	2.95 ms	49.0 ms	17× faster
First forecast in a new process	175 ms	1.46 s	8.3× faster
Save model to JSON	31 µs	703 µs	23× faster
Load model from JSON	137 µs	3.77 ms	28× faster
Load and predict in a new process	160 ms	1.42 s	8.9× faster

Settings

The case declaration both libraries ran.

{
  "kind": "map",
  "comparison": {
    "kind": "equivalent-objective",
    "evidenceId": "flat-map-mixed-v1"
  },
  "configuration": {
    "growth": "flat",
    "scaling": "absmax",
    "seasonalityMode": "additive",
    "holidaysMode": "additive",
    "changepoints": {
      "mode": "auto",
      "count": 25,
      "range": 0.8
    },
    "changepointPriorScale": 0.05,
    "seasonalities": [
      {
        "name": "weekly-custom",
        "periodDays": 7,
        "fourierOrder": 2,
        "priorScale": 10,
        "conditionName": "active",
        "mode": "multiplicative"
      }
    ],
    "events": [
      {
        "name": "campaign",
        "date": "2020-01-22",
        "lowerWindowDays": 0,
        "upperWindowDays": 0,
        "priorScale": 10
      }
    ],
    "regressors": [
      {
        "name": "promotion",
        "priorScale": 10,
        "standardization": "never"
      }
    ]
  },
  "pythonOptimizer": {
    "algorithm": "Auto",
    "maxIterations": 10000,
    "newtonFallback": true,
    "sigFigs": 12
  }
}