Are you wondering whether you should use Facebook’s Prophet forecasting model for your next data science project? Or maybe you are interested in hearing about the differences between Prophet and other time series forecasting models? Well either way, you are in the right place!
In this article, we tell you everything you need to know to determine when to use Prophet. We start out by talking about what kind of data Prophet is intended to be used for. After that, we discuss some of the main advantages and disadvantages of Prophet. Finally, we provide specific examples of use cases where you should and should not use Prophet.
What data should you use Facebook Prophet for?
What types of datasets should you use Facebook’s Prophet model on? In general, the Prophet model was designed to be used on time series data. Time series data is data that consists of many repeated observations on the same outcome variable.
Datasets that are used for time series modeling differ from datasets that are used for traditional supervised learning problems in that they do not need to include covariates that are associated with the outcome variable. While traditional supervised learning models predict future values of the outcome variable based on the values of covariates, time series models predict future values of the outcome variable based on previous values of the outcome variable.
Advantages and disadvantages of Facebook Prophet
Advantages of Facebook Prophet
What are some of the main advantages of Facebook’s Prophet model? Here are some advantages that Prophet has compared to other time series forecasting models.
- Can account for mean shifts. One of the main advantages of the Prophet model is that it is designed to be used in cases where there are mean shifts or disruptions in the time series data. That means that Prophet is a good model to reach for when there are mean shifts in your data, such as cases where there was a major product launch in the middle of your dataset.
- Can handle multiple seasonality. Another advantage of Prophet is that it can be used in situations where you have multiple different seasonal periods in your data. That means it is great for data that has daily, weekly, and yearly seasonality, for example.
- Designed for beginners. Another advantage of Prophet is that it was designed specifically for beginners who do not have a lot of experience with time series forecasting. That means that it is a good option for less technical teams that do not have a lot of experience with statistical modeling or time series forecasting.
- Robust to outliers. Another advantage of the Prophet model is that it is relatively robust to outliers. This is a quality that is somewhat rare among time series forecasting models, so it is certainly a notable characteristic.
- Robust to missing data. Another advantage of the Prophet model is that it is robust to missing data. That means that it is a good option to turn to if you have many missing values scattered throughout your dataset.
- Can incorporate covariates for holidays and events. Another advantage of the Prophet model is that it can incorporate covariates to represent the effects of holidays. This is a particularly important characteristic if you operate in an industry where the behavior of your customers is affected by holidays or special events.
- Does not require data to be stationary. The Prophet model takes a different approach to modeling time series data and treats the problem as more of a curve fitting problem. That means that it does not require data to be stationary. That being said, it often performs better on data that is stationary.
- Can handle non-linearity in trend. Another advantage of the Prophet model is that it can be used to model situations where there is non-linearity in the trend of the data. This means that it can be used to model situations where the relationship between a given data point and the data point that comes after it are not strictly linear.
- Interpretable decomposition. Another advantage of the Prophet model is that it provides an interpretable decomposition of the data that is observed or forecasted into seasonal, trend, and holiday components. This is useful if you want to understand why you are seeing certain patterns on certain days.
- Relatively computationally efficient. Another advantage of Prophet is that it tends to be relatively computationally efficient. Compared to other time series forecasting methods, it tends to be able to produce forecasts relatively quickly.
Disadvantages of Facebook Prophet
What are some of the main disadvantages of Facebook’s Prophet model? Here are some of the main disadvantages that Prophet has compared to other time series models.
- Subpar predictive performance. The main disadvantage of the Prophet model is that it tends to have subpar predictive performance when compared to classical time series models. That means that it is not a great option for people who are operating in situations where small increases in predictive performance can lead to large increases in business impact.
- Only appropriate for univariate time series. Another disadvantage of Prophet is that it can only be used for univariate time series models. That means that it is not ideal for situations where you have multiple correlated time series that you want to forecast jointly.
- Can’t incorporate all covariates. Another disadvantage of Facebook Prophet is that it is only designed to be able to handle covariates that represent holidays. It is not designed to handle other covariates that do not represent holidays or special events.
When to use Facebook Prophet
When should you use Facebook’s Prophet model rather than another time series forecasting model? Here are some examples of situations where you should use Facebook Prophet.
- When your data has large mean shifts. The Prophet model is specifically designed to be able to handle situations where there are large means shifts or disruptions in the data that is being forecasted. If you are building time series models using data that has large mean shifts then the prophet model might be a good option to look into.
- When you do not have experience with time series forecasting. The Prophet model was specifically designed with beginners in mind. The parameters were specified in a way such that they are easily interpretable for practitioners who do not have a lot of experience with time series forecasting. That means that Prophet might be a good model to reach for if you do not have anyone on your team who is experienced with statistical modeling or time series modeling.
When not to use Facebook Prophet
When should you avoid using Facebook’s Prophet model for time series forecasting? Here are some examples of situations where you should avoid using Facebook Prophet.
- When peak predictive performance is important. There are many situations where the Prophet forecasting models do not have as strong predictive performance as even simple time series models like ARIMA models and exponential smoothing models. If you are operating in a situation where small increases in predictive performance serve to generate large increases in business value, you may be better off using another method that requires more specialized skill.
Related articles
Time series models
- When to use exponential smoothing models
- When to use ARIMA models
- When to use TBATS
- When to use Fourier ARIMA models
Are you trying to figure out which machine learning model is best for your next data science project? Check out our comprehensive guide on how to choose the right machine learning model.