Course contentsShow
← Data Science
Lesson 795 of 2,14520. Time Series ModelsFree lesson

Seasonal ARIMA (SARIMA) Structure

Introducing seasonal autoregressive, differencing, and moving average components with notation (p,d,q)(P,D,Q)_s.

Seasonal ARIMA (SARIMA) Structure

What you'll learn: How SARIMA extends ARIMA by adding seasonal components to capture repeating patterns at fixed intervals.

The Core Idea

You've learned that ARIMA models capture non-seasonal patterns using three components: p (autoregression), d (differencing), and q (moving average). But what happens when your data has strong seasonal patterns—like monthly sales spikes every December or quarterly business cycles?

SARIMA (Seasonal ARIMA) adds a second layer of similar components that operate specifically on the seasonal lags. Instead of just looking at yesterday's value to predict today, SARIMA also looks at "last year's today" (or whatever your seasonal period is).

The SARIMA Notation: (p,d,q)(P,D,Q)_s

A SARIMA model is written as (p, d, q)(P, D, Q)_s, where:

  • (p, d, q) = your familiar non-seasonal ARIMA components
  • (P, D, Q) = the seasonal components:
    • P = seasonal autoregressive order
    • D = seasonal differencing order
    • Q = seasonal moving average order
  • s = the seasonal period (e.g., 12 for monthly data with yearly seasonality, 4 for quarterly data)

Think of it like this: The first set of parentheses handles short-term dependencies (recent observations), while the second set handles dependencies that repeat at seasonal intervals.

For example, ARIMA(1,1,1)(1,1,1)₁₂ means:

  • One non-seasonal AR term, one first-difference, one non-seasonal MA term
  • Plus one seasonal AR term, one seasonal difference, one seasonal MA term
  • With a 12-month seasonal cycle

Key Takeaway: SARIMA extends ARIMA with seasonal components (P,D,Q) that capture patterns repeating at fixed intervals s, giving you the full notation (p,d,q)(P,D,Q)_s to model both short-term and seasonal dependencies.