📘 How to Use Facebook Prophet for Time Series Forecasting

✅ What is Prophet?

Facebook Prophet is an open-source forecasting tool created by Facebook (Meta) designed for easy and accurate time series forecasting, especially for:

Daily or seasonal data

Business metrics (sales, traffic, etc.)

Irregular holidays or seasonality

It is robust to missing data and handles outliers well.

🛠️ 1. Installation

bash

Copy

Edit

pip install prophet

Note: If you have trouble installing Prophet, you may need pystan or cmdstanpy as a backend.

📁 2. Preparing Your Data

Prophet requires a DataFrame with two columns:

ds: Date column (datetime format)

y: The value you want to forecast

Example:

python

Copy

Edit

import pandas as pd

# Sample data

df = pd.read_csv('your_data.csv')

# Rename columns to 'ds' and 'y'

df.rename(columns={'date': 'ds', 'value': 'y'}, inplace=True)

df['ds'] = pd.to_datetime(df['ds'])

📈 3. Fit the Prophet Model

python

Copy

Edit

from prophet import Prophet

# Create the model

model = Prophet()

# Fit to your data

model.fit(df)

📅 4. Make Future Predictions

Create a DataFrame for Future Dates:

python

Copy

Edit

future = model.make_future_dataframe(periods=30)  # forecast 30 days ahead

Forecast:

python

Copy

Edit

forecast = model.predict(future)

📊 5. Visualize the Forecast

python

Copy

Edit

model.plot(forecast)

You can also plot the forecast components (trend, seasonality, holidays):

python

Copy

Edit

model.plot_components(forecast)

⚙️ 6. Add Seasonalities or Holidays (Optional)

Add Custom Seasonality:

python

Copy

Edit

model.add_seasonality(name='monthly', period=30.5, fourier_order=5)

Add Holidays:

python

Copy

Edit

holidays = pd.DataFrame({

  'holiday': 'event_name',

  'ds': pd.to_datetime(['2025-12-25', '2026-01-01']),

  'lower_window': 0,

  'upper_window': 1

})

model = Prophet(holidays=holidays)

✅ 7. Evaluate the Model (Optional)

You can split your data into training and testing sets and compare predictions with actual values.

python

Copy

Edit

from sklearn.metrics import mean_absolute_error

actual = df['y'][-30:].values

predicted = forecast['yhat'][-30:].values

mae = mean_absolute_error(actual, predicted)

print(f'MAE: {mae}')

📌 Key Features of Prophet

Feature Supported?

Trend Modeling ✅ Yes

Seasonality ✅ Yes

Holidays/Events ✅ Yes

Missing Data ✅ Handled

Outlier Robustness ✅ Yes

🧠 When to Use Prophet

Use Prophet when:

You have daily/weekly/monthly time series data.

Your data has strong seasonal patterns.

You want a simple and interpretable model.

You need quick and reasonably accurate forecasts.

🔚 Conclusion

Facebook Prophet is a great tool for:

Quick implementation

Business and financial forecasting

Handling real-world challenges like holidays and seasonality

Learn Data Science Course in Hyderabad

Read More

Common Time Series Forecasting Methods: ARIMA vs. LSTM

Introduction to Time Series Analysis in Data Science

12. Time Series Analysis and Forecasting

How to Use SHAP and LIME for Model Interpretability

Visit Our Quality Thought Training Institute in Hyderabad

Get Directions