Predicting from a known straight line

match ← All linear growth benchmarks

No fitting: both libraries evaluate the same stored line on 64 dates. Checks the prediction math on its own.

Input

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

Output

Largest difference: 1.4e-14 (< 0.0001% of the input's range).

Every forecast row from both libraries.

Date	Effect Prophet	Python Prophet	Difference
2020-09-13	36.1216	36.1216	7.1e-15
2020-09-14	36.2431	36.2431	7.1e-15
2020-09-15	36.3647	36.3647	0
2020-09-16	36.4863	36.4863	0
2020-09-17	36.6078	36.6078	1.4e-14
2020-09-18	36.7294	36.7294	7.1e-15
2020-09-19	36.8510	36.8510	0
2020-09-20	36.9725	36.9725	0
2020-09-21	37.0941	37.0941	7.1e-15
2020-09-22	37.2157	37.2157	7.1e-15
2020-09-23	37.3373	37.3373	0
2020-09-24	37.4588	37.4588	7.1e-15
2020-09-25	37.5804	37.5804	0
2020-09-26	37.7020	37.7020	7.1e-15
2020-09-27	37.8235	37.8235	7.1e-15
2020-09-28	37.9451	37.9451	7.1e-15
2020-09-29	38.0667	38.0667	0
2020-09-30	38.1882	38.1882	7.1e-15
2020-10-01	38.3098	38.3098	0
2020-10-02	38.4314	38.4314	7.1e-15
2020-10-03	38.5529	38.5529	0
2020-10-04	38.6745	38.6745	7.1e-15
2020-10-05	38.7961	38.7961	0
2020-10-06	38.9176	38.9176	7.1e-15
2020-10-07	39.0392	39.0392	7.1e-15
2020-10-08	39.1608	39.1608	7.1e-15
2020-10-09	39.2824	39.2824	0
2020-10-10	39.4039	39.4039	7.1e-15
2020-10-11	39.5255	39.5255	0
2020-10-12	39.6471	39.6471	7.1e-15
2020-10-13	39.7686	39.7686	0
2020-10-14	39.8902	39.8902	7.1e-15
2020-10-15	40.0118	40.0118	0
2020-10-16	40.1333	40.1333	0
2020-10-17	40.2549	40.2549	7.1e-15
2020-10-18	40.3765	40.3765	7.1e-15
2020-10-19	40.4980	40.4980	0
2020-10-20	40.6196	40.6196	7.1e-15
2020-10-21	40.7412	40.7412	0
2020-10-22	40.8627	40.8627	7.1e-15
2020-10-23	40.9843	40.9843	0
2020-10-24	41.1059	41.1059	7.1e-15
2020-10-25	41.2275	41.2275	0
2020-10-26	41.3490	41.3490	7.1e-15
2020-10-27	41.4706	41.4706	7.1e-15
2020-10-28	41.5922	41.5922	7.1e-15
2020-10-29	41.7137	41.7137	0
2020-10-30	41.8353	41.8353	7.1e-15
2020-10-31	41.9569	41.9569	0
2020-11-01	42.0784	42.0784	7.1e-15
2020-11-02	42.2000	42.2000	7.1e-15
2020-11-03	42.3216	42.3216	7.1e-15
2020-11-04	42.4431	42.4431	0
2020-11-05	42.5647	42.5647	7.1e-15
2020-11-06	42.6863	42.6863	0
2020-11-07	42.8078	42.8078	7.1e-15
2020-11-08	42.9294	42.9294	0
2020-11-09	43.0510	43.0510	7.1e-15
2020-11-10	43.1725	43.1725	0
2020-11-11	43.2941	43.2941	7.1e-15
2020-11-12	43.4157	43.4157	7.1e-15
2020-11-13	43.5373	43.5373	7.1e-15
2020-11-14	43.6588	43.6588	0
2020-11-15	43.7804	43.7804	7.1e-15

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Convert input rows	39 µs	562 µs	14× faster
Predict	304 µs	2.67 ms	8.8× faster

Settings

The case declaration both libraries ran.

{
  "kind": "fixed-linear-prediction",
  "comparison": {
    "kind": "equivalent-equation",
    "evidenceId": "fixed-linear-equation-v1"
  },
  "parameters": {
    "intercept": 5,
    "slope": 31,
    "timeOrigin": 1577836800000,
    "timeScale": 22032000000
  }
}