---
url: https://jer-k.github.io/effect-prophet/benchmarks/flat/flat-negative-minmax.md
---

# Negative level, minmax scaling

&#x20;[← All flat growth benchmarks](/benchmarks/flat)

The same negative rows with minmax scaling.

## Input

[`benchmark/inputs/v1/flat-negative.json`](https://github.com/jer-k/effect-prophet/blob/main/benchmark/inputs/v1/flat-negative.json): 96 history rows and 24 forecast rows.

## Output

Largest difference: **7.2e-8** (< 0.0001% of the input's range).

Every forecast row from both libraries.

| Date | Effect Prophet | Python Prophet | Difference |
| --- | ---: | ---: | ---: |
| 2020-04-06 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-07 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-08 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-09 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-10 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-11 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-12 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-13 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-14 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-15 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-16 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-17 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-18 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-19 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-20 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-21 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-22 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-23 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-24 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-25 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-26 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-27 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-28 | -34.9985 | -34.9985 | 7.2e-8 |
| 2020-04-29 | -34.9985 | -34.9985 | 7.2e-8 |

## Timing

Median of all measured repetitions.

| Step | Effect Prophet | Python Prophet | Effect Prophet is |
| --- | ---: | ---: | --- |
| Fit | 3.90 ms | 20.1 ms | 5.2× faster |
| Predict | 259 µs | 2.76 ms | 11× faster |
| Save model to JSON | 18 µs | 469 µs | 26× faster |
| Load model from JSON | 152 µs | 2.11 ms | 14× faster |
| Load and predict in a new process | 163 ms | 1.45 s | 8.9× faster |

## Settings

The case declaration both libraries ran.

```json
{
  "kind": "map",
  "comparison": {
    "kind": "equivalent-objective",
    "evidenceId": "flat-growth-reduced-v1"
  },
  "configuration": {
    "growth": "flat",
    "scaling": "minmax",
    "seasonalityMode": "additive",
    "holidaysMode": "additive",
    "changepoints": {
      "mode": "auto",
      "count": 25,
      "range": 0.8
    },
    "changepointPriorScale": 0.05,
    "seasonalities": [],
    "events": [],
    "regressors": []
  },
  "pythonOptimizer": {
    "algorithm": "Auto",
    "maxIterations": 10000,
    "newtonFallback": true,
    "sigFigs": 12
  }
}
```
