Data Science: Time Series Forecasting with Facebook Prophet in Python

In this self-paced course, you will learn how to use Facebook Prophet to do Time Series Analysis and Forecasting. You will learn about how Prophet works under the hood (i.e. what are its modeling assumptions?) and the Prophet API (i.e. how to write the code).

This course is a practice-oriented course, demonstrating how to prepare your data for Prophet, fit a model and use it to forecast, analyze the results, and evaluate the model's predictions. We will apply Prophet to a variety of datasets, including store sales and stock prices. The course includes video presentations, coding lessons, hands-on exercises, and links to further resources.

This course is intended for:

• Anyone interested in data science and machine learning
• Students and professionals who want to apply time series analysis to their own data

Suggested prerequisites:

• Decent Python programming skill
• Familiarity with Pandas and Dataframes
• Experience with scikit-learn is useful but not necessary

In this course, we will cover:

• how to prepare your data (a Pandas dataframe) for Facebook Prophet
• how to fit a Prophet model to a time series
• how to make a forecast using Prophet
• how to use Prophet to plot the model's in-sample predictions and forecast (with prediction intervals)
• how to plot the components of the fitted model (trend, error, and seasonal components)
• how to model holidays and exogenous regressors
• how to evaluate your model with forecasting metrics (MSE, RMSE, MAE, MAPE, sMAPE)
• how to perform cross-validation (walk-forward validation) with Prophet
• baselines, the random walk hypothesis, and the naive forecast
• how to do changepoint detection with Prophet
• how to model multiplicative seasonality with Prophet
• how to deal with outliers and missing data
• how to deal with non-daily (e.g. monthly) data
• how to use Prophet to predict stock prices