Model Based Approach

The Model Based Approach to Time Series forecasting fits a mathematical model - built on explicit statistical assumptions - to describe how the data behaves over time.

The Core Idea

# A model based approach assumes the data follows a specific
# structure, and estimates the parameters of that structure.
#
# Example: ARIMA assumes the series can be described using
# its own past values and past forecast errors.

A Common Example: ARIMA

from statsmodels.tsa.arima.model import ARIMA

model = ARIMA(df["Sales"], order=(1, 1, 1))
fit = model.fit()

forecast = fit.forecast(steps=3)
print(forecast)

Advantages

Model based approaches are interpretable - you can examine the model's parameters to understand relationships like trend strength or how much past values influence the future - and they work well with limited data.

Limitations

These models rely on assumptions (like stationarity) that real-world data doesn't always satisfy, and they can struggle to capture complex, non-linear patterns.

Before fitting a model like ARIMA, it's common practice to check and, if needed, transform the series to make it stationary - a step often done using differencing.

Coming Up Next