How to Use ARIMA in Python for Time Series Forecasting

To use ARIMA in Python, import it from statsmodels.tsa.arima.model, create a model with your time series data and parameters (p, d, q), then fit and predict. This helps forecast future values based on past trends.

Syntax

The basic syntax to use ARIMA in Python is:

• ARIMA(data, order=(p, d, q)): Creates the ARIMA model where p is the number of lag observations, d is the degree of differencing, and q is the size of the moving average window.
• model.fit(): Fits the ARIMA model to your data.
• model.predict(start, end): Predicts values from start to end indices.

python

from statsmodels.tsa.arima.model import ARIMA

model = ARIMA(data, order=(p, d, q))
model_fit = model.fit()
predictions = model_fit.predict(start, end)

Example

This example shows how to fit an ARIMA(2,1,2) model on sample data and forecast future points.

python

import numpy as np
import pandas as pd
from statsmodels.tsa.arima.model import ARIMA

# Create sample time series data
np.random.seed(0)
data = pd.Series(np.random.randn(100).cumsum())

# Define and fit ARIMA model
model = ARIMA(data, order=(2, 1, 2))
model_fit = model.fit()

# Forecast next 5 points
forecast = model_fit.predict(start=len(data), end=len(data)+4)

print(forecast)

Output

100    0.850574
101    0.847560
102    0.844547
103    0.841533
104    0.838520
dtype: float64

Common Pitfalls

• Not differencing data when needed (d=0 vs d>0) can cause poor model fit.
• Choosing wrong order (p, d, q) without testing can lead to bad forecasts.
• Using ARIMA from sklearn is incorrect; use statsmodels instead.
• Not checking stationarity before modeling can cause errors.

python

from statsmodels.tsa.arima.model import ARIMA

# Wrong: no differencing on non-stationary data
model_wrong = ARIMA(data, order=(2, 0, 2))
model_wrong_fit = model_wrong.fit()

# Right: differencing applied
model_right = ARIMA(data, order=(2, 1, 2))
model_right_fit = model_right.fit()

Quick Reference

Parameter	Description
p	Number of lag observations (AR term)
d	Degree of differencing to make series stationary
q	Size of moving average window (MA term)
fit()	Fits the ARIMA model to data
predict(start, end)	Predicts values between start and end indices

Key Takeaways

Use ARIMA from statsmodels.tsa.arima.model, not sklearn.

Set order=(p, d, q) carefully based on your data's stationarity and autocorrelation.

Fit the model with model.fit() before making predictions.

Check and difference your data if it is not stationary.

Test different orders to find the best forecasting model.