Weekly pattern, automatic changepoints

match ← All logistic growth benchmarks

The same history with two automatic changepoints instead of a hand-picked one.

Input

benchmark/inputs/v1/logistic-implicit-floor.json: 96 history rows and 24 forecast rows.

Output

Largest difference: 2.4e-11 (< 0.0001% of the input's range).

Every forecast row from both libraries.

Date	Effect Prophet	Python Prophet	Difference
2020-04-06	64.9080	64.9080	2.4e-11
2020-04-07	65.3515	65.3515	2.3e-11
2020-04-08	65.9703	65.9703	2.3e-11
2020-04-09	66.6561	66.6561	2.2e-11
2020-04-10	67.1311	67.1311	2.1e-11
2020-04-11	67.2247	67.2247	2.0e-11
2020-04-12	67.1798	67.1798	2.0e-11
2020-04-13	67.3238	67.3238	2.0e-11
2020-04-14	67.7093	67.7093	1.9e-11
2020-04-15	68.2719	68.2719	1.8e-11
2020-04-16	68.9030	68.9030	1.8e-11
2020-04-17	69.3249	69.3249	1.7e-11
2020-04-18	69.3671	69.3671	1.6e-11
2020-04-19	69.2723	69.2723	1.6e-11
2020-04-20	69.3679	69.3679	1.6e-11
2020-04-21	69.7068	69.7068	1.5e-11
2020-04-22	70.2240	70.2240	1.5e-11
2020-04-23	70.8115	70.8115	1.4e-11
2020-04-24	71.1911	71.1911	1.4e-11
2020-04-25	71.1924	71.1924	1.3e-11
2020-04-26	71.0583	71.0583	1.3e-11
2020-04-27	71.1159	71.1159	1.2e-11
2020-04-28	71.4181	71.4181	1.2e-11
2020-04-29	71.9000	71.9000	1.2e-11

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Convert input rows	19 µs	494 µs	26× faster
Fit	19.2 ms	30.6 ms	1.6× faster
Predict	237 µs	3.35 ms	14× faster
Fit and predict	18.9 ms	32.6 ms	1.7× faster
Convert, fit and predict	18.7 ms	34.0 ms	1.8× faster
First forecast in a new process	184 ms	1.45 s	7.9× faster
Save model to JSON	36 µs	500 µs	14× faster
Load model from JSON	103 µs	2.33 ms	23× faster
Load and predict in a new process	155 ms	1.42 s	9.1× faster

Settings

The case declaration both libraries ran.

{
  "kind": "map",
  "comparison": {
    "kind": "equivalent-objective",
    "evidenceId": "logistic-map-v1"
  },
  "configuration": {
    "growth": "logistic",
    "scaling": "minmax",
    "seasonalityMode": "additive",
    "holidaysMode": "additive",
    "changepoints": {
      "mode": "auto",
      "count": 2,
      "range": 0.8
    },
    "changepointPriorScale": 0.2,
    "seasonalities": [
      {
        "name": "weekly-custom",
        "periodDays": 7,
        "fourierOrder": 2,
        "priorScale": 10
      }
    ],
    "events": [],
    "regressors": []
  },
  "pythonOptimizer": {
    "algorithm": "Auto",
    "maxIterations": 10000,
    "newtonFallback": true,
    "sigFigs": 12
  }
}