Zansa › Answers › Forecasting and data science
What is overfitting and how can I tell if my forecast model has it?
Overfitting is when a model learns the noise and quirks of its training data rather than the real pattern, so it looks excellent on past data but performs poorly on new data. The tell-tale sign is a big gap between accuracy on training data and accuracy on held-out or backtest data.
Complex models with many features and little data overfit easily.
Ways to reduce it:
- Evaluate only on data the model has not seen, split by time.
- Prefer simpler models when results are similar.
- Limit the number of inputs.
- Use regularisation or tree depth limits in machine-learning methods.
- Compare with a baseline.
If adding complexity improves training error but not backtest error, you are fitting noise.
Want this applied to your business? Zansa builds dashboards and forecasts for small and mid-sized companies worldwide. Start with a free intro call.
Book a call