Calculator Input
Enter a time series, choose initialization, set smoothing controls, and generate fitted values plus future forecasts.
Example Data Table
Use this small trend-oriented series to test the model and compare fitted values against actual observations.
| Period | Observed Value | Comment |
|---|---|---|
| 1 | 120 | Starting baseline |
| 2 | 128 | Moderate increase |
| 3 | 133 | Trend continues |
| 4 | 141 | Stable acceleration |
| 5 | 150 | Higher demand |
| 6 | 158 | Growth persists |
| 7 | 165 | Smoother gain |
| 8 | 173 | Latest known point |
Formula Used
Holt linear trend forecasting smooths both the current level and the ongoing slope of a time series without seasonality.
How to Use This Calculator
- Paste at least four observations into the time series box, using commas, spaces, or line breaks.
- Choose the forecast horizon to control how many future periods you want predicted.
- Select an initialization method based on how you want the first level and trend states estimated.
- Either enter alpha and beta manually or enable auto-optimization for a grid search.
- Pick your preferred optimization criterion, such as RMSE or MAE, then set the grid step.
- Use the zero-actual checkbox when your data may contain zeros that would distort MAPE.
- Press Calculate Forecast to show results above the form, including metrics and fitted history.
- Download the report as CSV or PDF after results appear on the page.
Frequently Asked Questions
1. What does Holt linear trend forecasting do?
It estimates a changing level and a changing slope in a non-seasonal time series. The method is useful when data shows momentum but no repeating seasonal cycle.
2. When should I use this instead of simple exponential smoothing?
Use Holt’s method when your data trends upward or downward over time. Simple exponential smoothing is better when the series fluctuates around a stable average without a persistent slope.
3. What do alpha and beta control?
Alpha controls how quickly the model reacts to new level changes. Beta controls how fast the slope adjusts when the trend strengthens, weakens, or reverses.
4. Why might auto-optimization help?
Auto-optimization tests many alpha and beta combinations and selects the pair that minimizes your chosen error measure. It can improve fit when manual smoothing choices are uncertain.
5. Why are some APE values missing?
APE and MAPE can become undefined when an actual value is zero. The calculator can exclude those zero points from MAPE so percentage errors stay interpretable.
6. What is the tracking signal for?
Tracking signal compares cumulative error with average absolute error. Large positive or negative values may suggest persistent under-forecasting or over-forecasting bias in the model.
7. Can this calculator model seasonality?
No. Holt linear trend handles level and trend only. If your data has repeating seasonal patterns, use Holt-Winters or another seasonal forecasting method instead.
8. How far ahead should I forecast?
Shorter horizons are usually more stable because uncertainty grows with each step ahead. Choose a horizon aligned with the decision window you actually need to plan.