ARIMA (AutoRegressive Integrated Moving Average)
A classic statistical model for time series forecasting that combines autoregression, differencing, and moving averages.
ARIMA combines autoregression, differencing, and moving averages – the classic baseline model for time series forecasting.
Explanation
ARIMA(p,d,q): p = AR order, d = differencing for stationarity, q = MA order. SARIMA extends with seasonal components.
Marketing Relevance
ARIMA remains the baseline for any forecasting comparison. Fast, interpretable, good for univariate time series.
Example
ARIMA(1,1,1) forecasts monthly website traffic trends after removing the upward trend through differencing.
Common Pitfalls
Only for univariate, linear patterns. SARIMA needed for seasonal data. Cannot capture non-linear relationships.
Origin & History
Box & Jenkins formalized ARIMA in 1970. Industry standard for decades, supplemented by ML but never replaced as baseline.
Comparisons & Differences
ARIMA (AutoRegressive Integrated Moving Average) vs. Prophet
ARIMA requires manual parameter selection; Prophet automates trend/season decomposition.
ARIMA (AutoRegressive Integrated Moving Average) vs. LSTM
ARIMA models linear relationships; LSTM can learn non-linear patterns but needs more data.