100 rows: just over the optimizer switch

match ← All logistic growth benchmarks

Exactly 100 rows, so both libraries switch to L-BFGS.

Input

benchmark/inputs/v2/logistic-basic-100.json: 100 history rows and 124 forecast rows.

Output

Largest difference: 4.7e-11 (< 0.0001% of the input's range).

Every forecast row from both libraries.

Date	Effect Prophet	Python Prophet	Difference
2020-01-01	14.1874	14.1874	2.8e-11
2020-01-02	14.6981	14.6981	2.9e-11
2020-01-03	15.2241	15.2241	2.9e-11
2020-01-04	15.7656	15.7656	3.0e-11
2020-01-05	16.3228	16.3228	3.1e-11
2020-01-06	16.8959	16.8959	3.2e-11
2020-01-07	17.4852	17.4852	3.2e-11
2020-01-08	18.0908	18.0908	3.3e-11
2020-01-09	18.7128	18.7128	3.4e-11
2020-01-10	19.3514	19.3514	3.4e-11
2020-01-11	20.0067	20.0067	3.5e-11
2020-01-12	20.6788	20.6788	3.6e-11
2020-01-13	21.3678	21.3678	3.6e-11
2020-01-14	22.0737	22.0737	3.7e-11
2020-01-15	22.7966	22.7966	3.8e-11
2020-01-16	23.5364	23.5364	3.8e-11
2020-01-17	24.2932	24.2932	3.9e-11
2020-01-18	25.0668	25.0668	4.0e-11
2020-01-19	25.8572	25.8572	4.0e-11
2020-01-20	26.6642	26.6642	4.1e-11
2020-01-21	27.4877	27.4877	4.1e-11
2020-01-22	28.3275	28.3275	4.2e-11
2020-01-23	29.1833	29.1833	4.2e-11
2020-01-24	30.0550	30.0550	4.3e-11
2020-01-25	30.9421	30.9421	4.3e-11
2020-01-26	31.8445	31.8445	4.4e-11
2020-01-27	32.7616	32.7616	4.4e-11
2020-01-28	33.6932	33.6932	4.5e-11
2020-01-29	34.6387	34.6387	4.5e-11
2020-01-30	35.5977	35.5977	4.5e-11
2020-01-31	36.5698	36.5698	4.6e-11
2020-02-01	37.5542	37.5542	4.6e-11
2020-02-02	38.5505	38.5505	4.6e-11
2020-02-03	39.5581	39.5581	4.6e-11
2020-02-04	40.5763	40.5763	4.6e-11
2020-02-05	41.6045	41.6045	4.7e-11
2020-02-06	42.6419	42.6419	4.7e-11
2020-02-07	43.6879	43.6879	4.7e-11
2020-02-08	44.7416	44.7416	4.7e-11
2020-02-09	45.8026	45.8026	4.7e-11
2020-02-10	46.8697	46.8697	4.7e-11
2020-02-11	47.9423	47.9423	4.6e-11
2020-02-12	49.0195	49.0195	4.6e-11
2020-02-13	50.1005	50.1005	4.6e-11
2020-02-14	51.1844	51.1844	4.6e-11
2020-02-15	52.2704	52.2704	4.6e-11
2020-02-16	53.3576	53.3576	4.5e-11
2020-02-17	54.4452	54.4452	4.5e-11
2020-02-18	55.5323	55.5323	4.5e-11
2020-02-19	56.6181	56.6181	4.4e-11
2020-02-20	57.7018	57.7018	4.4e-11
2020-02-21	58.7824	58.7824	4.3e-11
2020-02-22	59.8592	59.8592	4.3e-11
2020-02-23	60.9314	60.9314	4.2e-11
2020-02-24	61.9982	61.9982	4.1e-11
2020-02-25	63.0588	63.0588	4.1e-11
2020-02-26	64.1124	64.1124	4.0e-11
2020-02-27	65.1584	65.1584	4.0e-11
2020-02-28	66.1961	66.1961	3.9e-11
2020-02-29	67.2247	67.2247	3.8e-11
2020-03-01	68.2436	68.2436	3.8e-11
2020-03-02	69.2523	69.2523	3.7e-11
2020-03-03	70.2501	70.2501	3.6e-11
2020-03-04	71.2365	71.2365	3.5e-11
2020-03-05	72.2110	72.2110	3.5e-11
2020-03-06	73.1731	73.1731	3.4e-11
2020-03-07	74.1223	74.1223	3.3e-11
2020-03-08	75.0583	75.0583	3.2e-11
2020-03-09	75.9806	75.9806	3.1e-11
2020-03-10	76.8889	76.8889	3.1e-11
2020-03-11	77.7830	77.7830	3.0e-11
2020-03-12	78.6624	78.6624	2.9e-11
2020-03-13	79.5271	79.5271	2.8e-11
2020-03-14	80.3767	80.3767	2.7e-11
2020-03-15	81.2112	81.2112	2.7e-11
2020-03-16	82.0303	82.0303	2.6e-11
2020-03-17	82.8339	82.8339	2.5e-11
2020-03-18	83.6220	83.6220	2.4e-11
2020-03-19	84.3945	84.3945	2.4e-11
2020-03-20	85.1514	85.1514	2.3e-11
2020-03-21	85.8927	85.8927	2.2e-11
2020-03-22	86.6184	86.6184	2.1e-11
2020-03-23	87.3285	87.3285	2.1e-11
2020-03-24	88.0231	88.0231	2.0e-11
2020-03-25	88.7024	88.7024	1.9e-11
2020-03-26	89.3663	89.3663	1.9e-11
2020-03-27	90.0151	90.0151	1.8e-11
2020-03-28	90.6489	90.6489	1.7e-11
2020-03-29	91.2679	91.2679	1.7e-11
2020-03-30	91.8721	91.8721	1.6e-11
2020-03-31	92.4619	92.4619	1.5e-11
2020-04-01	93.0375	93.0375	1.5e-11
2020-04-02	93.5989	93.5989	1.4e-11
2020-04-03	94.1465	94.1465	1.4e-11
2020-04-04	94.6805	94.6805	1.3e-11
2020-04-05	95.2011	95.2011	1.3e-11
2020-04-06	95.7086	95.7086	1.2e-11
2020-04-07	96.2033	96.2033	1.2e-11
2020-04-08	96.6853	96.6853	1.1e-11
2020-04-09	97.1550	97.1550	1.1e-11
2020-04-10	97.6126	97.6126	1.0e-11
2020-04-11	98.0583	98.0583	9.8e-12
2020-04-12	98.4926	98.4926	9.3e-12
2020-04-13	98.9155	98.9155	9.0e-12
2020-04-14	99.3275	99.3275	8.6e-12
2020-04-15	99.7287	99.7287	8.2e-12
2020-04-16	100.1194	100.1194	7.8e-12
2020-04-17	100.5000	100.5000	7.5e-12
2020-04-18	100.8706	100.8706	7.2e-12
2020-04-19	101.2316	101.2316	6.8e-12
2020-04-20	101.5831	101.5831	6.5e-12
2020-04-21	101.9255	101.9255	6.2e-12
2020-04-22	102.2591	102.2591	5.9e-12
2020-04-23	102.5840	102.5840	5.6e-12
2020-04-24	102.9005	102.9005	5.4e-12
2020-04-25	103.2089	103.2089	5.2e-12
2020-04-26	103.5094	103.5094	4.9e-12
2020-04-27	103.8022	103.8022	4.7e-12
2020-04-28	104.0876	104.0876	4.4e-12
2020-04-29	104.3658	104.3658	4.2e-12
2020-04-30	104.6371	104.6371	4.0e-12
2020-05-01	104.9015	104.9015	3.8e-12
2020-05-02	105.1595	105.1595	3.6e-12
2020-05-03	105.4111	105.4111	3.4e-12

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Fit	1.85 ms	7.74 ms	4.2× faster
Predict	819 µs	3.29 ms	4.0× faster
Save model to JSON	27 µs	473 µs	18× faster
Load model from JSON	157 µs	2.31 ms	15× faster
Load and predict in a new process	153 ms	1.57 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": [],
    "events": [],
    "regressors": []
  },
  "pythonOptimizer": {
    "algorithm": "Auto",
    "maxIterations": 10000,
    "newtonFallback": true,
    "sigFigs": 12
  }
}