Additive effects, minmax scaling

match ← All flat growth benchmarks

The mixed-effects rows with every effect additive: a switchable weekly pattern, an event and a regressor.

Input

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

Output

Largest difference: 2.3e-11 (< 0.0001% of the input's range).

Every forecast row from both libraries.

Date	Effect Prophet	Python Prophet	Difference
2020-04-06	21.3492	21.3492	8.6e-12
2020-04-07	24.0640	24.0640	2.0e-11
2020-04-08	36.2142	36.2142	2.1e-11
2020-04-09	45.9466	45.9466	1.4e-11
2020-04-10	35.0026	35.0026	7.7e-12
2020-04-11	41.0756	41.0756	7.6e-12
2020-04-12	28.9242	28.9242	3.7e-12
2020-04-13	21.3492	21.3492	8.6e-12
2020-04-14	25.2655	25.2655	1.9e-11
2020-04-15	35.0026	35.0026	7.7e-12
2020-04-16	45.9466	45.9466	1.4e-11
2020-04-17	48.6457	48.6457	4.3e-14
2020-04-18	41.0756	41.0756	7.6e-12
2020-04-19	28.9242	28.9242	3.7e-12
2020-04-20	36.2040	36.2040	6.4e-12
2020-04-21	24.0640	24.0640	2.0e-11
2020-04-22	35.0127	35.0127	2.3e-11
2020-04-23	45.9466	45.9466	1.4e-11
2020-04-24	48.6457	48.6457	5.0e-14
2020-04-25	35.0026	35.0026	7.7e-12
2020-04-26	30.1256	30.1256	5.0e-12
2020-04-27	21.3492	21.3492	8.6e-12
2020-04-28	24.0640	24.0640	2.0e-11
2020-04-29	35.0127	35.0127	2.3e-11

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Convert input rows	58 µs	518 µs	9.0× faster
Fit	5.48 ms	22.3 ms	4.1× faster
Predict	309 µs	4.64 ms	15× faster
Fit and predict	4.82 ms	27.5 ms	5.7× faster
Convert, fit and predict	4.52 ms	28.5 ms	6.3× faster
First forecast in a new process	177 ms	1.46 s	8.3× faster
Save model to JSON	35 µs	738 µs	21× faster
Load model from JSON	130 µs	3.83 ms	29× faster
Load and predict in a new process	159 ms	1.47 s	9.2× faster

Settings

The case declaration both libraries ran.

{
  "kind": "map",
  "comparison": {
    "kind": "equivalent-objective",
    "evidenceId": "flat-map-additive-v1"
  },
  "configuration": {
    "growth": "flat",
    "scaling": "minmax",
    "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"
      }
    ],
    "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
  }
}