ARIMA time series forecast helps you look into the future by identifying patterns in your historical data. Use this method when you have a sequence of values measured over consistent time intervals and want to project what happens next, such as estimating upcoming sales volume or future operational load.
Understanding Your Forecast
ARIMA stands for AutoRegressive Integrated Moving Average. It works by analyzing the relationship between current values and their past states (autoregression) while accounting for the errors and trends found in your previous data points.
When you run this tool, it breaks down your time series to identify the 'order' of your data. This tells you how many past steps the model needs to look back to provide the most accurate forecast for the future.
Automated Model Selection
You do not need to manually guess the model order. The tool automatically evaluates a range of possibilities and chooses the one that best fits your specific dataset. It evaluates the model quality using AIC and BIC metrics, which penalize overly complex models to prevent them from simply 'memorizing' noise in your data.
If your data is too sparse or irregular, the tool will provide a clear warning or fall back to a standard model (1,1,1), ensuring you always receive a result rather than a silent failure.
Diagnostic Checks
To ensure your results are trustworthy, the tool performs residual diagnostics. It checks whether the leftovers (residuals) are effectively white noise—meaning the model has successfully extracted all the predictable structure from your data.
If the Ljung-Box test indicates that significant patterns remain in the residuals, the tool flags this for you. This allows you to verify that the forecast is based on meaningful trends rather than random fluctuations.
Clear and Deterministic Results
Every forecast includes the point estimate (the most likely future value) and upper/lower bounds. Because the tool is deterministic, running the same data through Lattice will consistently produce the same result, making your analysis easy to replicate and defend in professional or academic contexts.
You receive a complete coefficient table, providing the statistical significance of each component used in your forecast, allowing you to see exactly which parts of your history are driving the projection.
1 · Intent → method
An LLM picks ts_arima_forecast from a fixed catalog.
2 · Method → numbers
Deterministic Python engine runs the math. Same input → same output.
3 · Numbers → plain language
A second LLM translates the result into your domain’s vocabulary.
How does Lattice decide which model settings to use?
The tool uses an automated grid search to test different combinations of patterns in your data. It selects the model with the lowest AIC/BIC score, which effectively balances the model's accuracy with its simplicity, ensuring you get the most reliable forecast without manual tuning.
What do the confidence intervals in the forecast mean?
The confidence intervals represent the range within which future values are likely to fall, based on your chosen confidence level (default 95%). If the range is narrow, the model is more certain; a wider range indicates higher variability or less predictable patterns in your historical data.
Tool input schema
Schema for ts_arima_forecast not exported yet (run pnpm export:registry).