Facebook Prophet: Outliers

from fbprophet import Prophet
import pandas as pd
import logging
logging.getLogger('fbprophet').setLevel(logging.ERROR)
import warnings
warnings.filterwarnings("ignore")

0.2s

Python

library(prophet)
library(ggplot2) # Required for Nextjournal plotting

0.2s

R

example_wp_log_peyton_manning.csv

There are two main ways that outliers can affect Prophet forecasts. Here we make a forecast on the logged Wikipedia visits to the R page from before, but with a block of bad data:

df <- read.csv(example_wp_log_peyton_manning.csv
)
m <- prophet(df)
future <- make_future_dataframe(m, periods = 1096)
forecast <- predict(m, future)
plot(m, forecast)

16.9s

R

df = pd.read_csv(example_wp_log_peyton_manning.csv
)
m = Prophet()
m.fit(df)
future = m.make_future_dataframe(periods=1096)
forecast = m.predict(future)
fig = m.plot(forecast)
fig

14.7s

Python

The trend forecast seems reasonable, but the uncertainty intervals seem way too wide. Prophet is able to handle the outliers in the history, but only by fitting them with trend changes. The uncertainty model then expects future trend changes of similar magnitude.

The best way to handle outliers is to remove them - Prophet has no problem with missing data. If you set their values to NA in the history but leave the dates in future, then Prophet will give you a prediction for their values.

outliers <- (as.Date(df$ds) > as.Date('2010-01-01')
             & as.Date(df$ds) < as.Date('2011-01-01'))
df$y[outliers] = NA
m <- prophet(df)
forecast <- predict(m, future)
plot(m, forecast)

14.0s

R

df.loc[(df['ds'] > '2010-01-01') & (df['ds'] < '2011-01-01'), 'y'] = None
model = Prophet().fit(df)
fig = model.plot(model.predict(future))
fig

15.0s

Python

In the above example the outliers messed up the uncertainty estimation but did not impact the main forecast yhat. This isn't always the case, as in this example with added outliers:

df <- read.csv(example_wp_log_peyton_manning.csv
)
m <- prophet(df)
future <- make_future_dataframe(m, periods = 1096)
forecast <- predict(m, future)
plot(m, forecast)

15.3s

R

df = pd.read_csv(example_wp_log_peyton_manning.csv
)
m = Prophet()
m.fit(df)
future = m.make_future_dataframe(periods=1096)
forecast = m.predict(future)
fig = m.plot(forecast)
fig

14.3s

Python

Here a group of extreme outliers in June 2015 mess up the seasonality estimate, so their effect reverberates into the future forever. Again the right approach is to remove them:

outliers <- (as.Date(df$ds) > as.Date('2015-06-01')
             & as.Date(df$ds) < as.Date('2015-06-30'))
df$y[outliers] = NA
m <- prophet(df)
forecast <- predict(m, future)
plot(m, forecast)

13.9s

R

df.loc[(df['ds'] > '2015-06-01') & (df['ds'] < '2015-06-30'), 'y'] = None
m = Prophet().fit(df)
fig = m.plot(m.predict(future))
fig

13.4s

Python