Built by Facebook's Core Data Science team, Prophet is designed for business forecasting at scale. It handles the messy reality of real-world data automatically.
1Curve Fitting vs. Stats
Traditional models like ARIMA rely on strict statistical assumptions and stationary data. Prophet takes a different approach: it treats forecasting as an Additive Regression problem. It combines a piecewise linear (or logistic) trend with multiple seasonal components and a list of holidays. This 'Curve Fitting' approach makes it much more resilient to missing data points and large outliers, which are common in business datasets.
2Automatic Components
One of Prophet's greatest strengths is that it automatically detects and models Daily, Weekly, and Yearly Seasonality. You don't need to difference your data or check for stationarity. If your data has at least two years of history, it will even model 'Yearly' patterns. You can also manually add specialized seasonalities, such as 'Monthly' or 'Bi-weekly,' to fit the specific needs of your business domain.
3Handling the Spikes
Business data is often dominated by Holidays and special events. Prophet allows you to provide a custom list of past and future holidays. The model then estimates the 'impact' of each holiday separately. This is much more effective than simple seasonal modeling, as holidays like Easter move to different dates every year. Prophet can even handle Change Points—sudden shifts in the trend caused by product launches or global events.
