Shuffled rows, automatic changepoints

match ← All linear growth benchmarks

Scrambled order with automatic changepoints, whose positions depend on the sorted rows.

Input

benchmark/inputs/v1/training-unsorted.json: 96 history rows and 8 forecast rows.

Output

Largest difference: 1.6e-9 (< 0.0001% of the input's range).

Every forecast row from both libraries.

Date	Effect Prophet	Python Prophet	Difference
2020-04-06	32.8119	32.8119	8.9e-10
2020-04-07	33.0424	33.0424	9.9e-10
2020-04-08	33.2730	33.2730	1.1e-9
2020-04-09	33.5036	33.5036	1.2e-9
2020-04-10	33.7341	33.7341	1.3e-9
2020-04-11	33.9647	33.9647	1.4e-9
2020-04-12	34.1953	34.1953	1.5e-9
2020-04-13	34.4258	34.4258	1.6e-9

Timing

Median of all measured repetitions.

Step	Effect Prophet	Python Prophet	Effect Prophet is
Fit	31.5 ms	107 ms	3.4× faster
Predict	151 µs	2.69 ms	18× faster
Save model to JSON	19 µs	404 µs	21× faster
Load model from JSON	169 µs	2.06 ms	12× faster
Load and predict in a new process	159 ms	1.42 s	8.9× faster

Settings

The case declaration both libraries ran.

{
  "kind": "map",
  "comparison": {
    "kind": "equivalent-objective",
    "evidenceId": "linear-output-first-v4"
  },
  "configuration": {
    "growth": "linear",
    "scaling": "absmax",
    "seasonalityMode": "additive",
    "holidaysMode": "additive",
    "changepoints": {
      "mode": "auto",
      "count": 25,
      "range": 0.8
    },
    "changepointPriorScale": 0.05,
    "seasonalities": [],
    "events": [],
    "regressors": []
  },
  "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
  }
}