Four regressors

match ← All linear growth benchmarks

256 daily rows with a yes/no regressor and three numeric ones, each standardized differently.

Input

benchmark/inputs/v1/map-regressors-medium.json: 256 history rows and 64 forecast rows.

Output

Largest difference: 0.0031 (0.0032% of the input's range).

Every forecast row from both libraries.

Date	Effect Prophet	Python Prophet	Difference
2020-09-13	107.1354	107.1342	0.0012
2020-09-14	107.5535	107.5522	0.0012
2020-09-15	107.9570	107.9558	0.0013
2020-09-16	108.3460	108.3447	0.0013
2020-09-17	107.8285	107.8272	0.0014
2020-09-18	110.0130	110.0115	0.0015
2020-09-19	110.3634	110.3619	0.0015
2020-09-20	110.7044	110.7029	0.0015
2020-09-21	109.6794	109.6779	0.0015
2020-09-22	110.0100	110.0085	0.0015
2020-09-23	110.3402	110.3387	0.0015
2020-09-24	110.6733	110.6718	0.0015
2020-09-25	111.4763	111.4748	0.0015
2020-09-26	111.8259	111.8244	0.0015
2020-09-27	114.0115	114.0099	0.0016
2020-09-28	114.3908	114.3891	0.0017
2020-09-29	115.2523	115.2507	0.0017
2020-09-30	112.4933	112.4917	0.0017
2020-10-01	112.9354	112.9337	0.0017
2020-10-02	113.4012	113.3995	0.0017
2020-10-03	114.3544	114.3527	0.0018
2020-10-04	114.8682	114.8664	0.0018
2020-10-05	115.4048	115.4030	0.0018
2020-10-06	117.7855	117.7836	0.0020
2020-10-07	118.8262	118.8243	0.0020
2020-10-08	119.4204	119.4184	0.0020
2020-10-09	118.2057	118.2038	0.0020
2020-10-10	118.8245	118.8225	0.0020
2020-10-11	119.9141	119.9120	0.0020
2020-10-12	120.5443	120.5422	0.0021
2020-10-13	119.8197	119.8175	0.0022
2020-10-14	120.4484	120.4462	0.0022
2020-10-15	123.3581	123.3557	0.0023
2020-10-16	123.9741	123.9718	0.0024
2020-10-17	124.5807	124.5783	0.0024
2020-10-18	123.3537	123.3514	0.0023
2020-10-19	124.4009	124.3986	0.0023
2020-10-20	124.9726	124.9702	0.0024
2020-10-21	125.5322	125.5298	0.0024
2020-10-22	126.0805	126.0781	0.0024
2020-10-23	127.0819	127.0795	0.0024
2020-10-24	129.4339	129.4315	0.0025
2020-10-25	129.9563	129.9538	0.0025
2020-10-26	129.1181	129.1155	0.0026
2020-10-27	128.2732	128.2707	0.0025
2020-10-28	128.7867	128.7842	0.0025
2020-10-29	129.3011	129.2986	0.0025
2020-10-30	129.8182	129.8157	0.0025
2020-10-31	130.8025	130.8000	0.0025
2020-11-01	131.3285	131.3259	0.0025
2020-11-02	133.6825	133.6799	0.0026
2020-11-03	134.2193	134.2166	0.0027
2020-11-04	135.2244	135.2217	0.0027
2020-11-05	133.9474	133.9448	0.0026
2020-11-06	134.4957	134.4931	0.0027
2020-11-07	135.0448	135.0421	0.0027
2020-11-08	134.7001	134.6973	0.0028
2020-11-09	135.2439	135.2410	0.0029
2020-11-10	135.7811	135.7782	0.0029
2020-11-11	138.1320	138.1290	0.0030
2020-11-12	139.1118	139.1087	0.0031
2020-11-13	139.6142	139.6111	0.0031
2020-11-14	138.2776	138.2745	0.0031
2020-11-15	138.7457	138.7426	0.0031

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Convert input rows	89 µs	660 µs	7.4× faster
Fit	2.96 ms	10.3 ms	3.5× faster
Predict	473 µs	3.52 ms	7.4× faster
Fit and predict	2.96 ms	13.6 ms	4.6× faster
Convert, fit and predict	2.69 ms	15.1 ms	5.6× faster
First forecast in a new process	170 ms	1.43 s	8.4× faster
Save model to JSON	24 µs	718 µs	30× faster
Load model from JSON	107 µs	2.97 ms	28× faster
Load and predict in a new process	165 ms	1.45 s	8.8× faster

Settings

The case declaration both libraries ran.

{
  "kind": "linear-map",
  "comparison": {
    "kind": "equivalent-objective",
    "evidenceId": "linear-output-first-v4"
  },
  "configuration": {
    "effectMap": "explicit",
    "changepoints": {
      "mode": "explicit",
      "timestamps": [
        "2020-03-21T00:00:00.000Z",
        "2020-06-29T00:00:00.000Z"
      ]
    },
    "changepointPriorScale": 0.2,
    "seasonalities": [],
    "events": [],
    "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
  }
}