Python Advanced - Time Series Analysis with Prophet

Performing time series analysis and forecasting using Prophet in Python

Prophet is an open-source forecasting tool developed by Facebook, designed for simplicity and flexibility in producing high-quality forecasts for time series data. This tutorial explores how to use Prophet for performing time series analysis and forecasting in Python.

Installing Prophet

To use Prophet, you need to install it using pip:

pip install prophet
            

Loading and Preparing Data

Here is an example of loading and preparing time series data using Pandas:

import pandas as pd

# Load the dataset
data = pd.read_csv('path/to/your/dataset.csv')

# Display the first few rows of the dataset
print(data.head())

# Rename columns to 'ds' and 'y' for Prophet
data.rename(columns={'date_column': 'ds', 'value_column': 'y'}, inplace=True)
            

Creating and Fitting a Prophet Model

Here is an example of creating and fitting a Prophet model:

from prophet import Prophet

# Create a Prophet model
model = Prophet()

# Fit the model to the data
model.fit(data)
            

Making Forecasts

Here is an example of making forecasts with a fitted Prophet model:

# Create a DataFrame for future dates
future = model.make_future_dataframe(periods=365)

# Make predictions
forecast = model.predict(future)

# Display the forecast
print(forecast[['ds', 'yhat', 'yhat_lower', 'yhat_upper']].tail())
            

Visualizing the Forecast

Here is an example of visualizing the forecast with Prophet:

import matplotlib.pyplot as plt

# Plot the forecast
fig1 = model.plot(forecast)
plt.show()

# Plot the forecast components
fig2 = model.plot_components(forecast)
plt.show()
            

Handling Seasonality and Holidays

Here is an example of modeling seasonality and adding holidays to a Prophet model:

# Create a Prophet model with yearly seasonality
model = Prophet(yearly_seasonality=True)

# Add country holidays
model.add_country_holidays(country_name='US')

# Fit the model to the data
model.fit(data)

# Create a DataFrame for future dates
future = model.make_future_dataframe(periods=365)

# Make predictions
forecast = model.predict(future)

# Plot the forecast
fig1 = model.plot(forecast)
plt.show()

# Plot the forecast components
fig2 = model.plot_components(forecast)
plt.show()
            

Custom Seasonalities

Here is an example of adding custom seasonalities to a Prophet model:

# Create a Prophet model
model = Prophet()

# Add a custom seasonality
model.add_seasonality(name='monthly', period=30.5, fourier_order=5)

# Fit the model to the data
model.fit(data)

# Create a DataFrame for future dates
future = model.make_future_dataframe(periods=365)

# Make predictions
forecast = model.predict(future)

# Plot the forecast
fig1 = model.plot(forecast)
plt.show()

# Plot the forecast components
fig2 = model.plot_components(forecast)
plt.show()
            

Evaluating Model Performance

Here is an example of evaluating the performance of a Prophet model:

from sklearn.metrics import mean_absolute_error

# Make predictions on the training data
train_pred = model.predict(data)

# Calculate the mean absolute error
mae = mean_absolute_error(data['y'], train_pred['yhat'])
print(f"Mean Absolute Error: {mae}")
            

Saving and Loading a Prophet Model

Here is an example of saving and loading a Prophet model:

import joblib

# Save the model
joblib.dump(model, 'prophet_model.pkl')

# Load the model
loaded_model = joblib.load('prophet_model.pkl')

# Verify the loaded model
future = loaded_model.make_future_dataframe(periods=365)
forecast = loaded_model.predict(future)
print(forecast[['ds', 'yhat']].tail())
            

Summary

In this tutorial, you learned about performing time series analysis and forecasting using Prophet in Python. Prophet is a powerful tool for time series forecasting developed by Facebook, designed to handle missing data, outliers, seasonality, and holidays. Understanding how to load and prepare data, create and fit a model, make forecasts, visualize results, handle seasonality and holidays, add custom seasonalities, evaluate model performance, and save/load models can help you leverage Prophet for high-quality time series analysis and forecasting.