

In [5]:

    
import pandas as pd
import numpy as np
from fbprophet import Prophet

DATA_HOME_DIR = '/data/airline'

The input to Prophet is always a dataframe with two columns: ds and y. The ds (datestamp) column must contain a date or datetime (either is fine). The y column must be numeric, and represents the measurement we wish to forecast.



In [60]:

    
df = pd.read_csv(DATA_HOME_DIR+'/international-airline-passengers.csv',
                 sep=';',
                 names=['ds', 'y'],
                 header=0,
                 parse_dates=[0],
                 nrows=144,
                )
df.head(3)
df.info()









    Out[60]:






  
    
      
      ds
      y
    
  
  
    
      0
      1949-01-01
      112
    
    
      1
      1949-02-01
      118
    
    
      2
      1949-03-01
      132
    
  








    



<class 'pandas.core.frame.DataFrame'>
RangeIndex: 144 entries, 0 to 143
Data columns (total 2 columns):
ds    144 non-null datetime64[ns]
y     144 non-null int64
dtypes: datetime64[ns](1), int64(1)
memory usage: 2.3 KB

It looks like we have a exponential growth trend in the data, so in order to accomodate for the linear fitting we take the log.



In [63]:

    
df['y'] = np.log(df['y'])

We only have monthly data, so certainly there will be no weekly seasonality in the date. Also forecasting must take this into account and choose the right frequency.



In [70]:

    
m = Prophet(weekly_seasonality=False)

m.fit(df)









    Out[70]:





<fbprophet.forecaster.Prophet at 0x7fef10349b00>

Predictions are then made on a dataframe with a column ds containing the dates for which a prediction is to be made. You can get a suitable dataframe that extends into the future a specified number of days using the helper method Prophet.make_future_dataframe. By default it will also include the dates from the history, so we will see the model fit as well.



In [71]:

    
future = m.make_future_dataframe(periods=36, freq='M')
future.tail()

The predict method will assign each row in future a predicted value which it names yhat. If you pass in historical dates, it will provide an in-sample fit. The forecast object here is a new dataframe that includes a column yhat with the forecast, as well as columns for components and uncertainty intervals.



In [72]:

    
forecast = m.predict(future)
forecast[['ds', 'yhat', 'yhat_lower', 'yhat_upper']].tail()



In [73]:

    
m.plot(forecast)









    Out[73]:

If you want to see the forecast components, you can use the Prophet.plot_components method. By default you’ll see the trend, yearly seasonality, and weekly seasonality of the time series. If you include holidays, you’ll see those here, too.



In [74]:

    
m.plot_components(forecast)









    Out[74]:



In [ ]:

	ds
176	1963-08-31
177	1963-09-30
178	1963-10-31
179	1963-11-30
180	1963-12-31

	ds	yhat	yhat_lower	yhat_upper
176	1963-08-31	6.597007	6.467691	6.725999
177	1963-09-30	6.423107	6.283125	6.561592
178	1963-10-31	6.291777	6.144299	6.434977
179	1963-11-30	6.443954	6.289768	6.594750
180	1963-12-31	6.473974	6.313121	6.625843