Mixed effects, absmax scaling

match ← All flat growth benchmarks

96 daily rows with a switchable multiplicative 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: 2.1e-7 (< 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	5.6e-8
2020-04-07	24.0640	24.0640	1.5e-7
2020-04-08	36.2142	36.2142	2.1e-7
2020-04-09	45.9466	45.9466	2.0e-7
2020-04-10	35.0026	35.0026	1.1e-7
2020-04-11	41.0756	41.0756	3.3e-8
2020-04-12	28.9241	28.9241	4.4e-9
2020-04-13	21.3492	21.3492	5.6e-8
2020-04-14	25.2655	25.2655	1.5e-7
2020-04-15	35.0026	35.0026	1.1e-7
2020-04-16	45.9466	45.9466	2.0e-7
2020-04-17	48.6457	48.6457	1.2e-7
2020-04-18	41.0756	41.0756	3.3e-8
2020-04-19	28.9241	28.9241	4.4e-9
2020-04-20	36.2041	36.2041	1.1e-7
2020-04-21	24.0640	24.0640	1.5e-7
2020-04-22	35.0127	35.0127	2.1e-7
2020-04-23	45.9466	45.9466	2.0e-7
2020-04-24	48.6457	48.6457	1.2e-7
2020-04-25	35.0026	35.0026	1.1e-7
2020-04-26	30.1256	30.1256	1.0e-9
2020-04-27	21.3492	21.3492	5.6e-8
2020-04-28	24.0640	24.0640	1.5e-7
2020-04-29	35.0127	35.0127	2.1e-7

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Fit	1.58 ms	7.92 ms	5.0× faster
Predict	469 µs	4.67 ms	10× faster
Save model to JSON	21 µs	651 µs	31× faster
Load model from JSON	226 µs	3.56 ms	16× faster
Load and predict in a new process	158 ms	1.41 s	9.0× faster

Settings

The case declaration both libraries ran.

{
  "kind": "map",
  "comparison": {
    "kind": "equivalent-objective",
    "evidenceId": "flat-growth-mixed-scaling-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"
      }
    ]
  },
  "effectOptimizer": {
    "algorithm": "lbfgs",
    "maxIterations": 10000,
    "fallback": "none",
    "lbfgs": {
      "historySize": 5,
      "initAlpha": 0.001,
      "tolObj": 1e-12,
      "tolRelObj": 10000,
      "tolGrad": 1e-8,
      "tolRelGrad": 10000000,
      "tolParam": 1e-8
    }
  },
  "pythonOptimizer": {
    "algorithm": "LBFGS",
    "maxIterations": 10000,
    "newtonFallback": false
  }
}