Additive effects, eight months, minmax scaling

match ← All flat growth benchmarks

The same additive features on 256 daily rows.

Input

benchmark/inputs/v1/flat-mixed-components-large.json: 256 history rows and 64 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
2020-09-13	28.9196	28.9196	2.3e-11
2020-09-14	21.3587	21.3587	1.4e-11
2020-09-15	24.0607	24.0607	3.0e-12
2020-09-16	34.9928	34.9928	1.4e-12
2020-09-17	36.2008	36.2008	1.0e-11
2020-09-18	48.6567	48.6567	1.5e-11
2020-09-19	41.0719	41.0719	2.3e-11
2020-09-20	28.9196	28.9196	2.3e-11
2020-09-21	21.3587	21.3587	1.4e-11
2020-09-22	35.0006	35.0006	1.1e-11
2020-09-23	36.1930	36.1930	2.7e-12
2020-09-24	45.9441	45.9441	3.7e-12
2020-09-25	48.6567	48.6567	1.5e-11
2020-09-26	41.0719	41.0719	2.3e-11
2020-09-27	35.0006	35.0006	1.1e-11
2020-09-28	21.3587	21.3587	1.4e-11
2020-09-29	25.2609	25.2609	1.8e-12
2020-09-30	34.9928	34.9928	1.4e-12
2020-10-01	45.9441	45.9441	3.7e-12
2020-10-02	35.0006	35.0006	1.1e-11
2020-10-03	41.0719	41.0719	2.3e-11
2020-10-04	28.9196	28.9196	2.3e-11
2020-10-05	22.5589	22.5589	1.3e-11
2020-10-06	24.0607	24.0607	3.0e-12
2020-10-07	35.0006	35.0006	1.1e-11
2020-10-08	45.9441	45.9441	3.7e-12
2020-10-09	48.6567	48.6567	1.5e-11
2020-10-10	41.0719	41.0719	2.3e-11
2020-10-11	30.1197	30.1197	2.1e-11
2020-10-12	35.0006	35.0006	1.1e-11
2020-10-13	24.0607	24.0607	3.0e-12
2020-10-14	34.9928	34.9928	1.4e-12
2020-10-15	45.9441	45.9441	3.7e-12
2020-10-16	48.6567	48.6567	1.5e-11
2020-10-17	36.2008	36.2008	1.0e-11
2020-10-18	28.9196	28.9196	2.3e-11
2020-10-19	21.3587	21.3587	1.4e-11
2020-10-20	24.0607	24.0607	3.0e-12
2020-10-21	34.9928	34.9928	1.4e-12
2020-10-22	35.0006	35.0006	1.1e-11
2020-10-23	49.8569	49.8569	1.3e-11
2020-10-24	41.0719	41.0719	2.3e-11
2020-10-25	28.9196	28.9196	2.3e-11
2020-10-26	21.3587	21.3587	1.4e-11
2020-10-27	35.0006	35.0006	1.1e-11
2020-10-28	34.9928	34.9928	1.4e-12
2020-10-29	47.1442	47.1442	2.5e-12
2020-10-30	48.6567	48.6567	1.5e-11
2020-10-31	41.0719	41.0719	2.3e-11
2020-11-01	35.0006	35.0006	1.1e-11
2020-11-02	21.3587	21.3587	1.4e-11
2020-11-03	24.0607	24.0607	3.0e-12
2020-11-04	36.1930	36.1930	2.7e-12
2020-11-05	45.9441	45.9441	3.7e-12
2020-11-06	35.0006	35.0006	1.1e-11
2020-11-07	41.0719	41.0719	2.3e-11
2020-11-08	28.9196	28.9196	2.3e-11
2020-11-09	21.3587	21.3587	1.4e-11
2020-11-10	25.2609	25.2609	1.8e-12
2020-11-11	35.0006	35.0006	1.1e-11
2020-11-12	45.9441	45.9441	3.7e-12
2020-11-13	48.6567	48.6567	1.5e-11
2020-11-14	41.0719	41.0719	2.3e-11
2020-11-15	28.9196	28.9196	2.3e-11

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Convert input rows	75 µs	643 µs	8.6× faster
Fit	2.16 ms	8.99 ms	4.2× faster
Predict	529 µs	4.67 ms	8.8× faster
Fit and predict	2.12 ms	13.3 ms	6.3× faster
Convert, fit and predict	2.16 ms	14.1 ms	6.5× faster
First forecast in a new process	170 ms	1.45 s	8.5× faster
Save model to JSON	27 µs	842 µs	32× faster
Load model from JSON	121 µs	3.91 ms	32× faster
Load and predict in a new process	159 ms	1.41 s	8.8× 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
  }
}