Model time series with practical ARIMA controls and diagnostics. Compare forecasts, residuals, and confidence bands. Turn raw observations into reliable planning signals for decisions.
This sample series shows steadily rising monthly demand and works well for testing ARIMA settings and forecast exports.
| Month | Observed Value |
|---|---|
| Jan | 120 |
| Feb | 128 |
| Mar | 133 |
| Apr | 140 |
| May | 149 |
| Jun | 158 |
| Jul | 166 |
| Aug | 172 |
| Sep | 181 |
| Oct | 190 |
| Nov | 198 |
| Dec | 207 |
y′t = yt − yt−1
d times removes trend and stabilizes the working series.
zt = c + φ1zt−1 + φ2zt−2 + ... + φpzt−p + et + θ1et−1 + θ2et−2 + ... + θqet−q
z is the differenced series, c is the constant, φ values are AR coefficients, and θ values are MA coefficients.
Forecast ± z × RMSE × √h
z comes from the selected confidence level, RMSE is the training error, and h is the forecast step.
Automatic mode uses conditional least squares for a fast transparent estimate. It is useful for planning and learning, while full production modeling may need deeper diagnostics.
ARIMA stands for AutoRegressive Integrated Moving Average. It models a series by combining past values, differencing, and past errors to produce forecasts.
The value p sets autoregressive lags, d sets how many times the series is differenced, and q sets moving average error lags.
Increase d when the original series has obvious trend and the model struggles with stability. Too much differencing can also distort useful structure.
Automatic mode estimates the constant and coefficients with conditional least squares on the differenced series, giving a quick practical model without outside libraries.
Yes. Manual mode lets you enter the constant, AR terms, and MA terms directly, which is useful for classroom examples and scenario testing.
They are approximate. This page scales training RMSE by the forecast step and confidence z score, so results are useful but not identical to full statistical software.
ARIMA works best with ordered numeric observations collected over equal time intervals, such as monthly demand, weekly traffic, or daily counts.
That usually means the chosen orders are not suitable, the series is too short, or the coefficients need adjustment. Try simpler settings first.
Important Note: All the Calculators listed in this site are for educational purpose only and we do not guarentee the accuracy of results. Please do consult with other sources as well.