Why More Samples Improve Accuracy
Industrial processes have natural variability. More samples help the model learn the full range of behavior:
- Normal operating range
- Rare events and process upsets
- Sensor drift over time
- Background noise patterns
- Non-linear zones at extremes
- Minimum: 30 samples
- Good: 50 samples
- Ideal: 70+ samples (especially for 6+ variable models)
Try it in the app
Try: In IoT Simulation mode, start a simulation and watch R² improve in real-time as more rows are collected. Notice how predictions stabilize beyond the 30-row mark.