Everything at once, two years

match flagged ← All linear growth benchmarks

768 daily rows with automatic changepoints, two seasonal patterns, six events and four regressors.

Input

benchmark/inputs/v1/map-mixed-features-automatic-large.json: 768 history rows and 128 forecast rows.

Output

Largest difference: 0.0069 (0.015% of the input's range).

Every forecast row from both libraries.

Date	Effect Prophet	Python Prophet	Difference
2022-02-07	46.5884	46.5852	0.0033
2022-02-08	47.3069	47.3037	0.0032
2022-02-09	47.5405	47.5370	0.0035
2022-02-10	47.9353	47.9317	0.0036
2022-02-11	49.5176	49.5144	0.0033
2022-02-12	49.5945	49.5911	0.0034
2022-02-13	50.5738	50.5691	0.0046
2022-02-14	49.6481	49.6433	0.0048
2022-02-15	50.3772	50.3722	0.0050
2022-02-16	49.3835	49.3795	0.0041
2022-02-17	49.4531	49.4490	0.0041
2022-02-18	49.2991	49.2950	0.0041
2022-02-19	48.8933	48.8895	0.0038
2022-02-20	48.1626	48.1589	0.0038
2022-02-21	46.1707	46.1668	0.0040
2022-02-22	48.6492	48.6441	0.0051
2022-02-23	50.3402	50.3350	0.0052
2022-02-24	51.0678	51.0625	0.0053
2022-02-25	49.7879	49.7838	0.0041
2022-02-26	50.1734	50.1694	0.0040
2022-02-27	49.9273	49.9235	0.0038
2022-02-28	49.1685	49.1649	0.0036
2022-03-01	49.7510	49.7472	0.0037
2022-03-02	50.8495	50.8457	0.0038
2022-03-03	53.1036	53.0990	0.0046
2022-03-04	51.8382	51.8335	0.0047
2022-03-05	51.9565	51.9517	0.0048
2022-03-06	49.4622	49.4585	0.0037
2022-03-07	49.1798	49.1760	0.0038
2022-03-08	49.9301	49.9262	0.0039
2022-03-09	51.1893	51.1854	0.0039
2022-03-10	55.5540	55.5492	0.0048
2022-03-11	52.1487	52.1443	0.0044
2022-03-12	55.1955	55.1885	0.0069
2022-03-13	57.6337	57.6278	0.0059
2022-03-14	52.3844	52.3795	0.0048
2022-03-15	51.3116	51.3077	0.0039
2022-03-16	52.1801	52.1759	0.0042
2022-03-17	50.8829	50.8787	0.0043
2022-03-18	50.6970	50.6926	0.0044
2022-03-19	50.6925	50.6881	0.0045
2022-03-20	49.4062	49.4017	0.0044
2022-03-21	50.5968	50.5915	0.0053
2022-03-22	51.0479	51.0425	0.0053
2022-03-23	51.1976	51.1920	0.0056
2022-03-24	49.6327	49.6282	0.0045
2022-03-25	49.8496	49.8451	0.0045
2022-03-26	50.0943	50.0898	0.0045
2022-03-27	49.8795	49.8750	0.0045
2022-03-28	49.2671	49.2628	0.0044
2022-03-29	49.9476	49.9434	0.0043
2022-03-30	50.5225	50.5176	0.0049
2022-03-31	48.7926	48.7889	0.0037
2022-04-01	52.2720	52.2667	0.0052
2022-04-02	50.2083	50.2045	0.0038
2022-04-03	49.2693	49.2656	0.0038
2022-04-04	48.7474	48.7436	0.0039
2022-04-05	49.3433	49.3393	0.0040
2022-04-06	50.5629	50.5587	0.0042
2022-04-07	51.1407	51.1364	0.0042
2022-04-08	53.7045	53.6992	0.0054
2022-04-09	53.9632	53.9580	0.0052
2022-04-10	52.4396	52.4342	0.0053
2022-04-11	49.7558	49.7516	0.0042
2022-04-12	50.2462	50.2424	0.0038
2022-04-13	51.3111	51.3071	0.0039
2022-04-14	51.6472	51.6431	0.0041
2022-04-15	51.7423	51.7380	0.0043
2022-04-16	52.0415	52.0370	0.0045
2022-04-17	52.8570	52.8513	0.0057
2022-04-18	51.7502	51.7444	0.0058
2022-04-19	51.9570	51.9511	0.0059
2022-04-20	51.7055	51.7009	0.0046
2022-04-21	52.1347	52.1301	0.0046
2022-04-22	51.2554	51.2506	0.0048
2022-04-23	51.2791	51.2743	0.0048
2022-04-24	50.9514	50.9465	0.0048
2022-04-25	48.9935	48.9887	0.0048
2022-04-26	51.7210	51.7150	0.0059
2022-04-27	53.1833	53.1774	0.0059
2022-04-28	54.2553	54.2494	0.0059
2022-04-29	52.8428	52.8382	0.0046
2022-04-30	53.0363	53.0318	0.0045
2022-05-01	52.3155	52.3110	0.0044
2022-05-02	51.7155	51.7114	0.0041
2022-05-03	52.2140	52.2098	0.0042
2022-05-04	53.3038	53.2995	0.0043
2022-05-05	55.6069	55.6013	0.0056
2022-05-06	56.3321	56.3264	0.0057
2022-05-07	56.6639	56.6582	0.0058
2022-05-08	52.9189	52.9144	0.0045
2022-05-09	52.3400	52.3356	0.0044
2022-05-10	52.6560	52.6515	0.0045
2022-05-11	53.7146	53.7100	0.0046
2022-05-12	55.0611	55.0568	0.0043
2022-05-13	55.2329	55.2286	0.0043
2022-05-14	57.4355	57.4300	0.0055
2022-05-15	56.5326	56.5269	0.0057
2022-05-16	55.5420	55.5361	0.0059
2022-05-17	54.0239	54.0189	0.0050
2022-05-18	55.1292	55.1240	0.0052
2022-05-19	55.0994	55.0942	0.0053
2022-05-20	55.1085	55.1035	0.0050
2022-05-21	53.7145	53.7096	0.0050
2022-05-22	52.1971	52.1920	0.0051
2022-05-23	52.9169	52.9107	0.0063
2022-05-24	53.3442	53.3379	0.0063
2022-05-25	54.4635	54.4571	0.0064
2022-05-26	53.5443	53.5389	0.0054
2022-05-27	53.9450	53.9396	0.0054
2022-05-28	54.2298	54.2246	0.0053
2022-05-29	53.5551	53.5500	0.0051
2022-05-30	53.0974	53.0925	0.0049
2022-05-31	53.5623	53.5574	0.0049
2022-06-01	56.2409	56.2351	0.0058
2022-06-02	56.4136	56.4078	0.0058
2022-06-03	55.3835	55.3778	0.0058
2022-06-04	53.5571	53.5523	0.0047
2022-06-05	52.7742	52.7694	0.0048
2022-06-06	52.0514	52.0465	0.0049
2022-06-07	53.1076	53.1026	0.0050
2022-06-08	54.3469	54.3420	0.0050
2022-06-09	54.1744	54.1693	0.0050
2022-06-10	56.0804	56.0742	0.0062
2022-06-11	57.3864	57.3806	0.0058
2022-06-12	56.5536	56.5478	0.0058
2022-06-13	53.8331	53.8284	0.0047
2022-06-14	54.2706	54.2657	0.0049

Internal fit differences

The forecasts match, so this case passes. Both libraries search for the best-fitting parameters and stop when they are close enough. On this input they stopped at slightly different points, so these internal numbers differ by more than the threshold where we take a closer look.

Check	Effect Prophet	Python Prophet	Threshold
Stationarity residual (lower is closer to the best fit)	176.1731	173.4560	0.01

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Convert input rows	182 µs	1.11 ms	6.1× faster
Fit	11.7 ms	37.3 ms	3.2× faster
Predict	957 µs	5.99 ms	6.3× faster
Fit and predict	12.9 ms	43.1 ms	3.3× faster
Convert, fit and predict	11.6 ms	44.5 ms	3.8× faster
First forecast in a new process	207 ms	1.45 s	7.0× faster
Save model to JSON	24 µs	1.71 ms	71× faster
Load model from JSON	269 µs	5.75 ms	21× faster
Load and predict in a new process	171 ms	1.41 s	8.3× faster

Settings

The case declaration both libraries ran.

{
  "kind": "linear-map",
  "comparison": {
    "kind": "equivalent-objective",
    "evidenceId": "linear-output-first-v4"
  },
  "configuration": {
    "effectMap": "omitted",
    "changepoints": {
      "mode": "auto",
      "count": 25,
      "range": 0.8
    },
    "changepointPriorScale": 0.05,
    "seasonalities": [
      {
        "name": "weekly-shared",
        "periodDays": 7,
        "fourierOrder": 2,
        "priorScale": 10
      },
      {
        "name": "monthly-active",
        "periodDays": 30.5,
        "fourierOrder": 3,
        "priorScale": 8,
        "conditionName": "active"
      }
    ],
    "events": [
      {
        "name": "release",
        "date": "2020-04-10",
        "lowerWindowDays": -1,
        "upperWindowDays": 2,
        "priorScale": 10
      },
      {
        "name": "release",
        "date": "2020-10-27",
        "lowerWindowDays": -1,
        "upperWindowDays": 2,
        "priorScale": 10
      },
      {
        "name": "release",
        "date": "2022-03-11",
        "lowerWindowDays": -1,
        "upperWindowDays": 2,
        "priorScale": 10
      },
      {
        "name": "campaign",
        "date": "2020-04-11",
        "lowerWindowDays": -1,
        "upperWindowDays": 0,
        "priorScale": 6
      },
      {
        "name": "campaign",
        "date": "2021-05-15",
        "lowerWindowDays": -1,
        "upperWindowDays": 0,
        "priorScale": 6
      },
      {
        "name": "campaign",
        "date": "2022-03-31",
        "lowerWindowDays": -1,
        "upperWindowDays": 0,
        "priorScale": 6
      }
    ],
    "regressors": [
      {
        "name": "binary-auto",
        "priorScale": 10,
        "standardization": "auto"
      },
      {
        "name": "numeric-auto",
        "priorScale": 10,
        "standardization": "auto"
      },
      {
        "name": "numeric-always",
        "priorScale": 7,
        "standardization": "always"
      },
      {
        "name": "numeric-never",
        "priorScale": 5,
        "standardization": "never"
      }
    ]
  },
  "effectOptimizer": {
    "algorithm": "auto",
    "maxIterations": 10000,
    "fallback": "newton",
    "lbfgs": {
      "historySize": 5,
      "initAlpha": 0.001,
      "tolObj": 1e-12,
      "tolRelObj": 10000,
      "tolGrad": 1e-8,
      "tolRelGrad": 10000000,
      "tolParam": 1e-8
    }
  },
  "pythonOptimizer": {
    "algorithm": "Auto",
    "maxIterations": 10000,
    "newtonFallback": true,
    "sigFigs": 12
  }
}