Events with windows

match ← All linear growth benchmarks

96 daily rows with six event dates, including days before and after each one.

Input

benchmark/inputs/v1/map-events-small.json: 96 history rows and 28 forecast rows.

Output

Largest difference: 7.3e-12 (< 0.0001% of the input's range).

Every forecast row from both libraries.

Date	Effect Prophet	Python Prophet	Difference
2020-04-06	45.3776	45.3776	7.3e-12
2020-04-07	45.7074	45.7074	7.3e-12
2020-04-08	46.0372	46.0372	7.3e-12
2020-04-09	46.3670	46.3670	7.2e-12
2020-04-10	46.6968	46.6968	7.2e-12
2020-04-11	47.0266	47.0266	7.2e-12
2020-04-12	47.3564	47.3564	7.2e-12
2020-04-13	47.6862	47.6862	7.2e-12
2020-04-14	48.0160	48.0160	7.2e-12
2020-04-15	50.9026	50.9026	5.5e-12
2020-04-16	51.5628	51.5628	4.8e-12
2020-04-17	51.9025	51.9025	6.8e-12
2020-04-18	51.7825	51.7825	6.1e-12
2020-04-19	49.6649	49.6649	7.1e-12
2020-04-20	49.9947	49.9947	7.0e-12
2020-04-21	50.3245	50.3245	7.0e-12
2020-04-22	48.4429	48.4429	5.7e-12
2020-04-23	48.7026	48.7026	7.3e-12
2020-04-24	51.3139	51.3139	7.0e-12
2020-04-25	51.6437	51.6437	7.0e-12
2020-04-26	51.9735	51.9735	6.9e-12
2020-04-27	52.3033	52.3033	6.9e-12
2020-04-28	52.6331	52.6331	6.9e-12
2020-04-29	52.9629	52.9629	6.9e-12
2020-04-30	53.2927	53.2927	6.9e-12
2020-05-01	53.6225	53.6225	6.8e-12
2020-05-02	53.9523	53.9523	6.8e-12
2020-05-03	54.2821	54.2821	6.8e-12

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Convert input rows	17 µs	479 µs	28× faster
Fit	7.18 ms	28.9 ms	4.0× faster
Predict	331 µs	4.21 ms	13× faster
Fit and predict	6.77 ms	33.1 ms	4.9× faster
Convert, fit and predict	6.68 ms	34.3 ms	5.1× faster
First forecast in a new process	177 ms	1.44 s	8.1× faster
Save model to JSON	26 µs	632 µs	24× faster
Load model from JSON	184 µs	3.41 ms	19× faster
Load and predict in a new process	158 ms	1.43 s	9.0× faster

Settings

The case declaration both libraries ran.

{
  "kind": "linear-map",
  "comparison": {
    "kind": "equivalent-objective",
    "evidenceId": "linear-output-first-v4"
  },
  "configuration": {
    "effectMap": "explicit",
    "changepoints": {
      "mode": "explicit",
      "timestamps": [
        "2020-02-10T00:00:00.000Z"
      ]
    },
    "changepointPriorScale": 0.2,
    "seasonalities": [],
    "events": [
      {
        "name": "release",
        "date": "2020-01-21",
        "lowerWindowDays": -1,
        "upperWindowDays": 2,
        "priorScale": 10
      },
      {
        "name": "release",
        "date": "2020-02-20",
        "lowerWindowDays": -1,
        "upperWindowDays": 2,
        "priorScale": 10
      },
      {
        "name": "release",
        "date": "2020-04-16",
        "lowerWindowDays": -1,
        "upperWindowDays": 2,
        "priorScale": 10
      },
      {
        "name": "campaign",
        "date": "2020-01-22",
        "lowerWindowDays": -1,
        "upperWindowDays": 0,
        "priorScale": 6
      },
      {
        "name": "campaign",
        "date": "2020-03-04",
        "lowerWindowDays": -1,
        "upperWindowDays": 0,
        "priorScale": 6
      },
      {
        "name": "campaign",
        "date": "2020-04-23",
        "lowerWindowDays": -1,
        "upperWindowDays": 0,
        "priorScale": 6
      }
    ],
    "regressors": []
  },
  "effectOptimizer": {
    "algorithm": "auto",
    "maxIterations": 10000,
    "fallback": "newton",
    "lbfgs": {
      "historySize": 5,
      "initAlpha": 0.001,
      "tolObj": 1e-12,
      "tolRelObj": 10000,
      "tolGrad": 1e-8,
      "tolRelGrad": 10000000,
      "tolParam": 1e-8
    }
  },
  "pythonOptimizer": {
    "algorithm": "Auto",
    "maxIterations": 10000,
    "newtonFallback": true,
    "sigFigs": 12
  }
}