Explicit floor, absmax scaling

match ← All logistic growth benchmarks

96 daily rows between a floor of about 12 and a ceiling of about 100.

Input

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

Output

Largest difference: 4.5e-11 (< 0.0001% of the input's range).

Every forecast row from both libraries.

Date	Effect Prophet	Python Prophet	Difference
2020-01-01	24.4855	24.4855	3.0e-11
2020-01-02	24.9722	24.9722	3.1e-11
2020-01-03	25.4734	25.4734	3.2e-11
2020-01-04	25.9892	25.9892	3.3e-11
2020-01-05	26.5199	26.5199	3.3e-11
2020-01-06	27.0657	27.0657	3.4e-11
2020-01-07	27.6267	27.6267	3.5e-11
2020-01-08	28.2032	28.2032	3.5e-11
2020-01-09	28.7953	28.7953	3.6e-11
2020-01-10	29.4030	29.4030	3.7e-11
2020-01-11	30.0266	30.0266	3.7e-11
2020-01-12	30.6661	30.6661	3.8e-11
2020-01-13	31.3215	31.3215	3.9e-11
2020-01-14	31.9929	31.9929	3.9e-11
2020-01-15	32.6803	32.6803	4.0e-11
2020-01-16	33.3836	33.3836	4.0e-11
2020-01-17	34.1028	34.1028	4.1e-11
2020-01-18	34.8378	34.8378	4.1e-11
2020-01-19	35.5885	35.5885	4.2e-11
2020-01-20	36.3548	36.3548	4.2e-11
2020-01-21	37.1364	37.1364	4.3e-11
2020-01-22	37.9331	37.9331	4.3e-11
2020-01-23	38.7446	38.7446	4.4e-11
2020-01-24	39.5707	39.5707	4.4e-11
2020-01-25	40.4111	40.4111	4.4e-11
2020-01-26	41.2653	41.2653	4.5e-11
2020-01-27	42.1329	42.1329	4.5e-11
2020-01-28	43.0136	43.0136	4.5e-11
2020-01-29	43.9068	43.9068	4.5e-11
2020-01-30	44.8121	44.8121	4.5e-11
2020-01-31	45.7288	45.7288	4.5e-11
2020-02-01	46.6563	46.6563	4.5e-11
2020-02-02	47.5942	47.5942	4.5e-11
2020-02-03	48.5417	48.5417	4.5e-11
2020-02-04	49.4981	49.4981	4.5e-11
2020-02-05	50.4628	50.4628	4.5e-11
2020-02-06	51.4350	51.4350	4.5e-11
2020-02-07	52.4139	52.4139	4.5e-11
2020-02-08	53.3989	53.3989	4.4e-11
2020-02-09	54.3894	54.3894	4.4e-11
2020-02-10	55.3842	55.3842	4.4e-11
2020-02-11	56.3827	56.3827	4.3e-11
2020-02-12	57.3839	57.3839	4.3e-11
2020-02-13	58.3870	58.3870	4.2e-11
2020-02-14	59.3911	59.3911	4.2e-11
2020-02-15	60.3955	60.3955	4.1e-11
2020-02-16	61.3992	61.3992	4.1e-11
2020-02-17	62.4014	62.4014	4.0e-11
2020-02-18	63.4013	63.4013	4.0e-11
2020-02-19	64.3981	64.3981	3.9e-11
2020-02-20	65.3910	65.3910	3.8e-11
2020-02-21	66.3791	66.3791	3.7e-11
2020-02-22	67.3617	67.3617	3.7e-11
2020-02-23	68.3381	68.3381	3.6e-11
2020-02-24	69.3074	69.3074	3.5e-11
2020-02-25	70.2691	70.2691	3.4e-11
2020-02-26	71.2223	71.2223	3.3e-11
2020-02-27	72.1666	72.1666	3.3e-11
2020-02-28	73.1012	73.1012	3.2e-11
2020-02-29	74.0256	74.0256	3.1e-11
2020-03-01	74.9391	74.9391	3.0e-11
2020-03-02	75.8414	75.8414	2.9e-11
2020-03-03	76.7319	76.7319	2.8e-11
2020-03-04	77.6101	77.6101	2.7e-11
2020-03-05	78.4757	78.4757	2.6e-11
2020-03-06	79.3282	79.3282	2.6e-11
2020-03-07	80.1674	80.1674	2.5e-11
2020-03-08	80.9928	80.9928	2.4e-11
2020-03-09	81.8044	81.8044	2.3e-11
2020-03-10	82.6017	82.6017	2.2e-11
2020-03-11	83.3846	83.3846	2.1e-11
2020-03-12	84.1529	84.1529	2.1e-11
2020-03-13	84.9066	84.9066	2.0e-11
2020-03-14	85.6454	85.6454	1.9e-11
2020-03-15	86.3694	86.3694	1.8e-11
2020-03-16	87.0784	87.0784	1.7e-11
2020-03-17	87.7725	87.7725	1.7e-11
2020-03-18	88.4516	88.4516	1.6e-11
2020-03-19	89.1159	89.1159	1.5e-11
2020-03-20	89.7653	89.7653	1.4e-11
2020-03-21	90.3999	90.3999	1.4e-11
2020-03-22	91.0199	91.0199	1.3e-11
2020-03-23	91.6253	91.6253	1.2e-11
2020-03-24	92.2163	92.2163	1.2e-11
2020-03-25	92.7931	92.7931	1.1e-11
2020-03-26	93.3558	93.3558	1.1e-11
2020-03-27	93.9047	93.9047	1.0e-11
2020-03-28	94.4398	94.4398	9.5e-12
2020-03-29	94.9615	94.9615	9.0e-12
2020-03-30	95.4699	95.4699	8.5e-12
2020-03-31	95.9653	95.9653	8.0e-12
2020-04-01	96.4480	96.4480	7.6e-12
2020-04-02	96.9181	96.9181	7.1e-12
2020-04-03	97.3759	97.3759	6.7e-12
2020-04-04	97.8217	97.8217	6.3e-12
2020-04-05	98.2558	98.2558	5.9e-12
2020-04-06	98.6784	98.6784	5.5e-12
2020-04-07	99.0897	99.0897	5.2e-12
2020-04-08	99.4901	99.4901	4.8e-12
2020-04-09	99.8799	99.8799	4.5e-12
2020-04-10	100.2592	100.2592	4.2e-12
2020-04-11	100.6283	100.6283	3.9e-12
2020-04-12	100.9876	100.9876	3.6e-12
2020-04-13	101.3373	101.3373	3.3e-12
2020-04-14	101.6777	101.6777	3.1e-12
2020-04-15	102.0090	102.0090	2.8e-12
2020-04-16	102.3315	102.3315	2.6e-12
2020-04-17	102.6454	102.6454	2.4e-12
2020-04-18	102.9510	102.9510	2.1e-12
2020-04-19	103.2485	103.2485	2.0e-12
2020-04-20	103.5383	103.5383	1.8e-12
2020-04-21	103.8204	103.8204	1.6e-12
2020-04-22	104.0953	104.0953	1.5e-12
2020-04-23	104.3630	104.3630	1.3e-12
2020-04-24	104.6239	104.6239	1.2e-12
2020-04-25	104.8781	104.8781	1.0e-12
2020-04-26	105.1259	105.1259	8.8e-13
2020-04-27	105.3675	105.3675	7.7e-13
2020-04-28	105.6031	105.6031	6.5e-13
2020-04-29	105.8329	105.8329	5.7e-13

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Fit	7.75 ms	13.8 ms	1.8× faster
Predict	883 µs	2.89 ms	3.3× faster
Save model to JSON	20 µs	470 µs	24× faster
Load model from JSON	160 µs	2.26 ms	14× faster
Load and predict in a new process	162 ms	1.44 s	8.9× 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
  }
}