One hand-picked changepoint, minmax 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. The rate may change on 8 February. Values are scaled between the lowest floor and the highest ceiling.

Input

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

Output

Largest difference: 3.9e-11 (< 0.0001% of the input's range).

Every forecast row from both libraries.

Date	Effect Prophet	Python Prophet	Difference
2020-01-01	17.9893	17.9893	1.3e-11
2020-01-02	18.2431	18.2431	1.3e-11
2020-01-03	18.5130	18.5130	1.4e-11
2020-01-04	18.8001	18.8001	1.5e-11
2020-01-05	19.1053	19.1053	1.5e-11
2020-01-06	19.4295	19.4295	1.6e-11
2020-01-07	19.7738	19.7738	1.7e-11
2020-01-08	20.1393	20.1393	1.8e-11
2020-01-09	20.5269	20.5269	1.9e-11
2020-01-10	20.9379	20.9379	1.9e-11
2020-01-11	21.3734	21.3734	2.0e-11
2020-01-12	21.8345	21.8345	2.1e-11
2020-01-13	22.3223	22.3223	2.2e-11
2020-01-14	22.8381	22.8381	2.3e-11
2020-01-15	23.3830	23.3830	2.4e-11
2020-01-16	23.9582	23.9582	2.5e-11
2020-01-17	24.5648	24.5648	2.6e-11
2020-01-18	25.2040	25.2040	2.7e-11
2020-01-19	25.8769	25.8769	2.8e-11
2020-01-20	26.5845	26.5845	2.9e-11
2020-01-21	27.3279	27.3279	3.0e-11
2020-01-22	28.1079	28.1079	3.0e-11
2020-01-23	28.9254	28.9254	3.1e-11
2020-01-24	29.7811	29.7811	3.2e-11
2020-01-25	30.6758	30.6758	3.3e-11
2020-01-26	31.6099	31.6099	3.4e-11
2020-01-27	32.5838	32.5838	3.5e-11
2020-01-28	33.5977	33.5977	3.5e-11
2020-01-29	34.6517	34.6517	3.6e-11
2020-01-30	35.7457	35.7457	3.7e-11
2020-01-31	36.8792	36.8792	3.7e-11
2020-02-01	38.0519	38.0519	3.8e-11
2020-02-02	39.2628	39.2628	3.8e-11
2020-02-03	40.5111	40.5111	3.8e-11
2020-02-04	41.7955	41.7955	3.9e-11
2020-02-05	43.1145	43.1145	3.9e-11
2020-02-06	44.4665	44.4665	3.9e-11
2020-02-07	45.8496	45.8496	3.9e-11
2020-02-08	47.2616	47.2616	3.8e-11
2020-02-09	48.2516	48.2516	3.8e-11
2020-02-10	49.2537	49.2537	3.7e-11
2020-02-11	50.2670	50.2670	3.7e-11
2020-02-12	51.2908	51.2908	3.6e-11
2020-02-13	52.3241	52.3241	3.5e-11
2020-02-14	53.3659	53.3659	3.5e-11
2020-02-15	54.4153	54.4153	3.4e-11
2020-02-16	55.4711	55.4711	3.3e-11
2020-02-17	56.5324	56.5324	3.2e-11
2020-02-18	57.5982	57.5982	3.1e-11
2020-02-19	58.6673	58.6673	3.0e-11
2020-02-20	59.7387	59.7387	2.9e-11
2020-02-21	60.8112	60.8112	2.8e-11
2020-02-22	61.8839	61.8839	2.7e-11
2020-02-23	62.9555	62.9555	2.6e-11
2020-02-24	64.0250	64.0250	2.5e-11
2020-02-25	65.0914	65.0914	2.4e-11
2020-02-26	66.1536	66.1536	2.2e-11
2020-02-27	67.2106	67.2106	2.1e-11
2020-02-28	68.2613	68.2613	2.0e-11
2020-02-29	69.3049	69.3049	1.9e-11
2020-03-01	70.3402	70.3402	1.8e-11
2020-03-02	71.3666	71.3666	1.7e-11
2020-03-03	72.3830	72.3830	1.5e-11
2020-03-04	73.3887	73.3887	1.4e-11
2020-03-05	74.3828	74.3828	1.3e-11
2020-03-06	75.3647	75.3647	1.2e-11
2020-03-07	76.3336	76.3336	1.1e-11
2020-03-08	77.2889	77.2889	9.8e-12
2020-03-09	78.2301	78.2301	8.8e-12
2020-03-10	79.1565	79.1565	7.7e-12
2020-03-11	80.0676	80.0676	6.7e-12
2020-03-12	80.9632	80.9632	5.8e-12
2020-03-13	81.8426	81.8426	4.8e-12
2020-03-14	82.7056	82.7056	3.9e-12
2020-03-15	83.5520	83.5520	3.0e-12
2020-03-16	84.3814	84.3814	2.2e-12
2020-03-17	85.1937	85.1937	1.4e-12
2020-03-18	85.9886	85.9886	6.3e-13
2020-03-19	86.7662	86.7662	1.1e-13
2020-03-20	87.5262	87.5262	7.7e-13
2020-03-21	88.2688	88.2688	1.4e-12
2020-03-22	88.9939	88.9939	2.1e-12
2020-03-23	89.7015	89.7015	2.6e-12
2020-03-24	90.3917	90.3917	3.2e-12
2020-03-25	91.0646	91.0646	3.7e-12
2020-03-26	91.7204	91.7204	4.2e-12
2020-03-27	92.3592	92.3592	4.6e-12
2020-03-28	92.9812	92.9812	5.0e-12
2020-03-29	93.5866	93.5866	5.4e-12
2020-03-30	94.1757	94.1757	5.8e-12
2020-03-31	94.7486	94.7486	6.1e-12
2020-04-01	95.3058	95.3058	6.4e-12
2020-04-02	95.8474	95.8474	6.6e-12
2020-04-03	96.3737	96.3737	6.8e-12
2020-04-04	96.8851	96.8851	7.0e-12
2020-04-05	97.3819	97.3819	7.2e-12
2020-04-06	97.8644	97.8644	7.4e-12
2020-04-07	98.3329	98.3329	7.5e-12
2020-04-08	98.7877	98.7877	7.6e-12
2020-04-09	99.2293	99.2293	7.7e-12
2020-04-10	99.6579	99.6579	7.8e-12
2020-04-11	100.0739	100.0739	7.9e-12
2020-04-12	100.4777	100.4777	7.9e-12
2020-04-13	100.8695	100.8695	7.9e-12
2020-04-14	101.2497	101.2497	7.9e-12
2020-04-15	101.6188	101.6188	7.9e-12
2020-04-16	101.9769	101.9769	7.9e-12
2020-04-17	102.3245	102.3245	7.9e-12
2020-04-18	102.6618	102.6618	7.8e-12
2020-04-19	102.9893	102.9893	7.8e-12
2020-04-20	103.3072	103.3072	7.7e-12
2020-04-21	103.6158	103.6158	7.6e-12
2020-04-22	103.9154	103.9154	7.6e-12
2020-04-23	104.2064	104.2064	7.5e-12
2020-04-24	104.4891	104.4891	7.4e-12
2020-04-25	104.7637	104.7637	7.3e-12
2020-04-26	105.0305	105.0305	7.2e-12
2020-04-27	105.2899	105.2899	7.0e-12
2020-04-28	105.5420	105.5420	7.0e-12
2020-04-29	105.7871	105.7871	6.8e-12

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Fit	16.8 ms	24.2 ms	1.4× faster
Predict	724 µs	2.87 ms	4.0× faster
Save model to JSON	28 µs	432 µs	15× faster
Load model from JSON	141 µs	2.17 ms	15× faster
Load and predict in a new process	157 ms	1.58 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": "minmax",
    "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
  }
}