Everything at once, absmax scaling

match ← All logistic growth benchmarks

A switchable weekly pattern, a campaign event and a multiplicative regressor between an explicit floor and ceiling.

Input

benchmark/inputs/v2/logistic-mixed-96.json: 96 history rows and 120 forecast rows.

Output

Largest difference: 4.3e-11 (< 0.0001% of the input's range).

Every forecast row from both libraries.

Date	Effect Prophet	Python Prophet	Difference
2020-01-01	24.5263	24.5263	3.5e-12
2020-01-02	25.6388	25.6388	3.0e-12
2020-01-03	27.2295	27.2295	2.7e-12
2020-01-04	26.3756	26.3756	2.3e-12
2020-01-05	26.2047	26.2047	2.0e-12
2020-01-06	27.1083	27.1083	1.5e-12
2020-01-07	27.0356	27.0356	1.3e-12
2020-01-08	28.2706	28.2706	6.3e-13
2020-01-09	30.5190	30.5190	6.8e-14
2020-01-10	30.2275	30.2275	6.7e-13
2020-01-11	30.0701	30.0701	1.1e-12
2020-01-12	30.3521	30.3521	1.6e-12
2020-01-13	30.5813	30.5813	2.2e-12
2020-01-14	31.4023	31.4023	2.8e-12
2020-01-15	33.9451	33.9451	3.8e-12
2020-01-16	33.4267	33.4267	4.4e-12
2020-01-17	34.9266	34.9266	5.3e-12
2020-01-18	35.2248	35.2248	6.0e-12
2020-01-19	35.2731	35.2731	6.7e-12
2020-01-20	35.6128	35.6128	7.4e-12
2020-01-21	40.6815	40.6815	9.1e-12
2020-01-22	39.8265	39.8265	5.8e-12
2020-01-23	41.3651	41.3651	1.2e-11
2020-01-24	40.3910	40.3910	1.1e-11
2020-01-25	40.7940	40.7940	1.2e-11
2020-01-26	41.3029	41.3029	1.3e-11
2020-01-27	42.9288	42.9288	1.4e-11
2020-01-28	42.4154	42.4154	1.5e-11
2020-01-29	43.9659	43.9659	1.6e-11
2020-01-30	45.4711	45.4711	1.7e-11
2020-01-31	45.7612	45.7612	1.8e-11
2020-02-01	47.0321	47.0321	1.9e-11
2020-02-02	49.0090	49.0090	2.1e-11
2020-02-03	47.7866	47.7866	2.1e-11
2020-02-04	48.8914	48.8914	2.2e-11
2020-02-05	50.4886	50.4886	2.3e-11
2020-02-06	52.0847	52.0847	2.5e-11
2020-02-07	53.2178	53.2178	2.6e-11
2020-02-08	55.7191	55.7191	2.8e-11
2020-02-09	54.0520	54.0520	2.7e-11
2020-02-10	55.4033	55.4033	2.9e-11
2020-02-11	55.7666	55.7666	2.9e-11
2020-02-12	57.4250	57.4250	3.0e-11
2020-02-13	59.0279	59.0279	3.2e-11
2020-02-14	62.3599	62.3599	3.4e-11
2020-02-15	60.4086	60.4086	3.3e-11
2020-02-16	61.0536	61.0536	3.4e-11
2020-02-17	61.6284	61.6284	3.5e-11
2020-02-18	62.7768	62.7768	3.5e-11
2020-02-19	64.4307	64.4307	3.6e-11
2020-02-20	67.7906	67.7906	3.8e-11
2020-02-21	67.1659	67.1659	3.8e-11
2020-02-22	67.7108	67.7108	3.8e-11
2020-02-23	67.9839	67.9839	3.9e-11
2020-02-24	68.5260	68.5260	3.9e-11
2020-02-25	70.2701	70.2701	4.0e-11
2020-02-26	73.8523	73.8523	4.2e-11
2020-02-27	72.7908	72.7908	4.1e-11
2020-02-28	73.8799	73.8799	4.1e-11
2020-02-29	74.3667	74.3667	4.1e-11
2020-03-01	74.9348	74.9348	4.2e-11
2020-03-02	75.0524	75.0524	4.2e-11
2020-03-03	78.8987	78.8987	4.3e-11
2020-03-04	77.6272	77.6272	4.2e-11
2020-03-05	79.0930	79.0930	4.2e-11
2020-03-06	79.3193	79.3193	4.2e-11
2020-03-07	80.5021	80.5021	4.2e-11
2020-03-08	80.6248	80.6248	4.2e-11
2020-03-09	84.0019	84.0019	4.3e-11
2020-03-10	81.9557	81.9557	4.1e-11
2020-03-11	83.3720	83.3720	4.1e-11
2020-03-12	84.7651	84.7651	4.1e-11
2020-03-13	85.6738	85.6738	4.1e-11
2020-03-14	85.9756	85.9756	4.1e-11
2020-03-15	89.1564	89.1564	4.2e-11
2020-03-16	87.0632	87.0632	4.0e-11
2020-03-17	87.1229	87.1229	3.9e-11
2020-03-18	88.4599	88.4599	3.9e-11
2020-03-19	89.7249	89.7249	3.9e-11
2020-03-20	90.5296	90.5296	3.9e-11
2020-03-21	93.6897	93.6897	3.9e-11
2020-03-22	90.6452	90.6452	3.7e-11
2020-03-23	90.8242	90.8242	3.7e-11
2020-03-24	91.5648	91.5648	3.6e-11
2020-03-25	92.7998	92.7998	3.6e-11
2020-03-26	93.3380	93.3380	3.5e-11
2020-03-27	98.1027	98.1027	3.6e-11
2020-03-28	94.7664	94.7664	3.4e-11
2020-03-29	94.5860	94.5860	3.3e-11
2020-03-30	94.6683	94.6683	3.3e-11
2020-03-31	95.9474	95.9474	3.2e-11
2020-04-01	96.4544	96.4544	3.2e-11
2020-04-02	101.0709	101.0709	3.2e-11
2020-04-03	98.1392	98.1392	3.1e-11
2020-04-04	98.1486	98.1486	3.0e-11
2020-04-05	98.2383	98.2383	2.9e-11
2020-04-06	97.8774	97.8774	2.8e-11
2020-04-07	98.4386	98.4386	2.8e-11
2020-04-08	103.1368	103.1368	2.8e-11
2020-04-09	100.4884	100.4884	2.7e-11
2020-04-10	100.2425	100.2425	2.6e-11
2020-04-11	100.9564	100.9564	2.6e-11
2020-04-12	100.6139	100.6139	2.5e-11
2020-04-13	100.5376	100.5376	2.4e-11
2020-04-14	104.7473	104.7473	2.4e-11
2020-04-15	104.1373	104.1373	2.3e-11
2020-04-16	104.7701	104.7701	1.9e-11
2020-04-17	105.3665	105.3665	2.4e-11
2020-04-18	103.2806	103.2806	2.1e-11
2020-04-19	102.8765	102.8765	2.1e-11
2020-04-20	107.3115	107.3115	2.1e-11
2020-04-21	103.1724	103.1724	1.9e-11
2020-04-22	104.1058	104.1058	1.9e-11
2020-04-23	104.9749	104.9749	1.9e-11
2020-04-24	105.3917	105.3917	1.8e-11
2020-04-25	104.8652	104.8652	1.8e-11
2020-04-26	108.6014	108.6014	1.8e-11
2020-04-27	104.5714	104.5714	1.6e-11
2020-04-28	104.9570	104.9570	1.6e-11
2020-04-29	105.8453	105.8453	1.6e-11

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Fit	29.0 ms	46.6 ms	1.6× faster
Predict	1.02 ms	5.61 ms	5.5× faster
Save model to JSON	21 µs	691 µs	32× faster
Load model from JSON	176 µs	4.05 ms	23× faster
Load and predict in a new process	158 ms	1.60 s	10× faster

Settings

The case declaration both libraries ran.

{
  "kind": "map",
  "comparison": {
    "kind": "equivalent-objective",
    "evidenceId": "logistic-python-reconciliation-v1"
  },
  "configuration": {
    "growth": "logistic",
    "scaling": "absmax",
    "seasonalityMode": "additive",
    "holidaysMode": "additive",
    "changepoints": {
      "mode": "explicit",
      "timestamps": [
        "2020-02-08T00:00:00.000Z"
      ]
    },
    "changepointPriorScale": 0.05,
    "seasonalities": [
      {
        "name": "weekly-custom",
        "periodDays": 7,
        "fourierOrder": 3,
        "priorScale": 10,
        "conditionName": "active"
      }
    ],
    "events": [
      {
        "name": "campaign",
        "date": "2020-01-22",
        "lowerWindowDays": -1,
        "upperWindowDays": 1,
        "priorScale": 10
      },
      {
        "name": "campaign",
        "date": "2020-04-16",
        "lowerWindowDays": -1,
        "upperWindowDays": 1,
        "priorScale": 10
      }
    ],
    "regressors": [
      {
        "name": "promotion",
        "priorScale": 10,
        "standardization": "never",
        "mode": "multiplicative"
      }
    ]
  },
  "pythonOptimizer": {
    "algorithm": "Auto",
    "maxIterations": 10000,
    "newtonFallback": true,
    "sigFigs": 12
  }
}