Three automatic changepoints, absmax scaling

match ← All logistic growth benchmarks

96 daily rows rising toward a ceiling of about 100, with a little noise. Prediction covers the history and 24 more days. Changepoints are spread over the first 80% of the history. Values are divided by their largest distance from the floor.

Input

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

Output

Largest difference: 1.3e-11 (< 0.0001% of the input's range).

Every forecast row from both libraries.

Date	Effect Prophet	Python Prophet	Difference
2020-01-01	14.1882	14.1882	2.9e-12
2020-01-02	14.7203	14.7203	2.8e-12
2020-01-03	15.2688	15.2688	2.7e-12
2020-01-04	15.8342	15.8342	2.6e-12
2020-01-05	16.4167	16.4167	2.5e-12
2020-01-06	17.0164	17.0164	2.4e-12
2020-01-07	17.6337	17.6337	2.3e-12
2020-01-08	18.2686	18.2686	2.2e-12
2020-01-09	18.9214	18.9214	2.0e-12
2020-01-10	19.5922	19.5922	1.9e-12
2020-01-11	20.2812	20.2812	1.7e-12
2020-01-12	20.9883	20.9883	1.6e-12
2020-01-13	21.7138	21.7138	1.4e-12
2020-01-14	22.4576	22.4576	1.2e-12
2020-01-15	23.2198	23.2198	1.0e-12
2020-01-16	24.0003	24.0003	7.9e-13
2020-01-17	24.7991	24.7991	5.6e-13
2020-01-18	25.6161	25.6161	3.3e-13
2020-01-19	26.4511	26.4511	9.6e-14
2020-01-20	27.3040	27.3040	1.6e-13
2020-01-21	28.1745	28.1745	4.3e-13
2020-01-22	29.0625	29.0625	7.0e-13
2020-01-23	29.9675	29.9675	9.7e-13
2020-01-24	30.8894	30.8894	1.3e-12
2020-01-25	31.8277	31.8277	1.6e-12
2020-01-26	32.7820	32.7820	1.9e-12
2020-01-27	33.7521	33.7521	2.2e-12
2020-01-28	34.7372	34.7372	2.5e-12
2020-01-29	35.7368	35.7368	2.9e-12
2020-01-30	36.7503	36.7503	3.2e-12
2020-01-31	37.7772	37.7772	3.5e-12
2020-02-01	38.8167	38.8167	3.9e-12
2020-02-02	39.8681	39.8681	4.2e-12
2020-02-03	40.9308	40.9308	4.6e-12
2020-02-04	42.0039	42.0039	4.9e-12
2020-02-05	43.0866	43.0866	5.3e-12
2020-02-06	44.1782	44.1782	5.7e-12
2020-02-07	45.2777	45.2777	6.0e-12
2020-02-08	46.3843	46.3843	6.4e-12
2020-02-09	47.4971	47.4971	6.7e-12
2020-02-10	48.6151	48.6151	7.1e-12
2020-02-11	49.7374	49.7374	7.4e-12
2020-02-12	50.8631	50.8631	7.8e-12
2020-02-13	51.9912	51.9912	8.1e-12
2020-02-14	53.1208	53.1208	8.4e-12
2020-02-15	54.2508	54.2508	8.8e-12
2020-02-16	55.3803	55.3803	9.1e-12
2020-02-17	56.5084	56.5084	9.4e-12
2020-02-18	57.6341	57.6341	9.7e-12
2020-02-19	58.7565	58.7565	9.9e-12
2020-02-20	59.8746	59.8746	1.0e-11
2020-02-21	60.9877	60.9877	1.0e-11
2020-02-22	62.0948	62.0948	1.1e-11
2020-02-23	63.1949	63.1949	1.1e-11
2020-02-24	64.2873	64.2873	1.1e-11
2020-02-25	65.3711	65.3711	1.1e-11
2020-02-26	66.4456	66.4456	1.2e-11
2020-02-27	67.5099	67.5099	1.2e-11
2020-02-28	68.5634	68.5634	1.2e-11
2020-02-29	69.6055	69.6055	1.2e-11
2020-03-01	70.6354	70.6354	1.2e-11
2020-03-02	71.6526	71.6526	1.2e-11
2020-03-03	72.6565	72.6565	1.2e-11
2020-03-04	73.6466	73.6466	1.3e-11
2020-03-05	74.6224	74.6224	1.3e-11
2020-03-06	75.5834	75.5834	1.3e-11
2020-03-07	76.5294	76.5294	1.3e-11
2020-03-08	77.4599	77.4599	1.3e-11
2020-03-09	78.3746	78.3746	1.3e-11
2020-03-10	79.2732	79.2732	1.3e-11
2020-03-11	80.1555	80.1555	1.3e-11
2020-03-12	81.0212	81.0212	1.3e-11
2020-03-13	81.8704	81.8704	1.3e-11
2020-03-14	82.7027	82.7027	1.3e-11
2020-03-15	83.5181	83.5181	1.3e-11
2020-03-16	84.3166	84.3166	1.2e-11
2020-03-17	85.0981	85.0981	1.2e-11
2020-03-18	85.8626	85.8626	1.2e-11
2020-03-19	86.6103	86.6103	1.2e-11
2020-03-20	87.3410	87.3410	1.2e-11
2020-03-21	88.0550	88.0550	1.2e-11
2020-03-22	88.7522	88.7522	1.2e-11
2020-03-23	89.4330	89.4330	1.2e-11
2020-03-24	90.0973	90.0973	1.2e-11
2020-03-25	90.7454	90.7454	1.1e-11
2020-03-26	91.3776	91.3776	1.1e-11
2020-03-27	91.9938	91.9938	1.1e-11
2020-03-28	92.5946	92.5946	1.1e-11
2020-03-29	93.1799	93.1799	1.1e-11
2020-03-30	93.7502	93.7502	1.1e-11
2020-03-31	94.3056	94.3056	1.0e-11
2020-04-01	94.8465	94.8465	1.0e-11
2020-04-02	95.3731	95.3731	1.0e-11
2020-04-03	95.8856	95.8856	9.8e-12
2020-04-04	96.3845	96.3845	9.6e-12
2020-04-05	96.8700	96.8700	9.4e-12
2020-04-06	97.3423	97.3423	9.3e-12
2020-04-07	97.8018	97.8018	9.0e-12
2020-04-08	98.2489	98.2489	8.9e-12
2020-04-09	98.6837	98.6837	8.7e-12
2020-04-10	99.1067	99.1067	8.5e-12
2020-04-11	99.5180	99.5180	8.3e-12
2020-04-12	99.9181	99.9181	8.1e-12
2020-04-13	100.3073	100.3073	7.9e-12
2020-04-14	100.6857	100.6857	7.8e-12
2020-04-15	101.0538	101.0538	7.5e-12
2020-04-16	101.4118	101.4118	7.4e-12
2020-04-17	101.7600	101.7600	7.2e-12
2020-04-18	102.0988	102.0988	7.0e-12
2020-04-19	102.4283	102.4283	6.8e-12
2020-04-20	102.7489	102.7489	6.7e-12
2020-04-21	103.0608	103.0608	6.5e-12
2020-04-22	103.3644	103.3644	6.4e-12
2020-04-23	103.6598	103.6598	6.2e-12
2020-04-24	103.9474	103.9474	6.0e-12
2020-04-25	104.2273	104.2273	5.8e-12
2020-04-26	104.4999	104.4999	5.7e-12
2020-04-27	104.7654	104.7654	5.5e-12
2020-04-28	105.0240	105.0240	5.4e-12
2020-04-29	105.2760	105.2760	5.2e-12

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Fit	19.0 ms	30.2 ms	1.6× faster
Predict	699 µs	3.53 ms	5.0× faster
Save model to JSON	28 µs	459 µs	17× faster
Load model from JSON	160 µs	2.29 ms	14× faster
Load and predict in a new process	160 ms	1.43 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": "auto",
      "count": 3,
      "range": 0.8
    },
    "changepointPriorScale": 0.05,
    "seasonalities": [],
    "events": [],
    "regressors": []
  },
  "pythonOptimizer": {
    "algorithm": "Auto",
    "maxIterations": 10000,
    "newtonFallback": true,
    "sigFigs": 12
  }
}