Forecasting with Facebook's Prophet

This notebook shows how to use the Prophet-algorithm developed by Facebook for forecasting time series.

# First install the required packages
!pip install pandas pystan matplotlib
!pip install fbprophet

import pandas as pd
import matplotlib.pyplot as plt
from fbprophet import Prophet
from fbprophet.plot import plot_yearly

Read the data

df = pd.read_csv("data/GlobalTemperatures.csv", parse_dates=[0])
df = df[["dt", "LandAverageTemperature"]]
df["LandAverageTemperature"] = df["LandAverageTemperature"].interpolate()

df.rename(columns={"dt":"ds", "LandAverageTemperature": "y"}, inplace=True)

Split data by percentage

split_percentage = 80
train_size = int(df.shape[0]*(split_percentage/100))
train, test = df[:train_size], df[train_size:]

Create model and predict future

m = Prophet()
m.fit(train)

future = m.make_future_dataframe(periods=len(test)+1, freq='m', include_history=False)
future.tail()

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

INFO:fbprophet:Disabling weekly seasonality. Run prophet with weekly_seasonality=True to override this.
INFO:fbprophet:Disabling daily seasonality. Run prophet with daily_seasonality=True to override this.
C:\ProgramData\Anaconda3\lib\site-packages\pystan\misc.py:399: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.
  elif np.issubdtype(np.asarray(v).dtype, float):

	ds	yhat	yhat_lower	yhat_upper
635	2015-08-31	12.797079	11.677862	14.044220
636	2015-09-30	10.403715	9.112078	11.612158
637	2015-10-31	6.404978	5.277403	7.581737
638	2015-11-30	4.220793	2.963257	5.403963
639	2015-12-31	3.097171	1.889019	4.350749

Plot the results

fig1 = m.plot(forecast)

fig2 = m.plot_components(forecast)

Especially the last trend plot shows how the average temperature rose the past years.

Split data by year

train_start_date = '1900-01-01'
train_end_date = '1999-12-01'
test_start_date = '2000-01-01'
test_end_date = '2015-12-01'

train = df[(df["ds"] &gt;= train_start_date) & (df["ds"] &lt;= train_end_date)]
test = df[(df["ds"] &gt;= test_start_date) & (df["ds"] &lt;= test_end_date)]

Create model and predict future

m = Prophet()
m.fit(train)

future = m.make_future_dataframe(periods=len(test)+1, freq='m', include_history=False)
future.tail()

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

INFO:fbprophet:Disabling weekly seasonality. Run prophet with weekly_seasonality=True to override this.
INFO:fbprophet:Disabling daily seasonality. Run prophet with daily_seasonality=True to override this.
C:\ProgramData\Anaconda3\lib\site-packages\pystan\misc.py:399: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.
  elif np.issubdtype(np.asarray(v).dtype, float):

	ds	yhat	yhat_lower	yhat_upper
188	2015-08-31	13.183082	12.762133	13.612796
189	2015-09-30	10.683178	10.259379	11.101987
190	2015-10-31	7.143251	6.748653	7.575937
191	2015-11-30	4.828418	4.415568	5.261541
192	2015-12-31	3.825875	3.425469	4.271024

Create model and predict future

fig1 = m.plot(forecast)

fig2 = m.plot_components(forecast)

The last trend plot shows again how the average temperature rose the past years.