Forecasting Made Easy with Facebook Prophet

Table of Contents

• Introduction to Facebook Prophet
  • How It works
• Why you should use Facebook Prophet
• Preparing environment for Forecasting
  • Installation of Facebook Prophet
  • Importing libraries
  • Data to Forecast
• Forcasting using Facebook Prophet
  • Example 1: Forecasting Electric production

Introduction to Facebook Prophet

Facebook Prophet is an open-source forecasting tool developed by Facebook. It can make accurate forecasts on time-series data. Even when dealing with seasonality, holidays and other patterns which are common in real world.

This is particularly useful for business forecasting such as predicting sales or stock demand.

How It works

It uses an additive model that combines the trend of data. By doing this, it achieves the overall upward and downward trend of the data over time.

It models periodic patterns like yearly, monthly and weekly fluctuation while also accounting for existing holidays in the data.

Why you should use Facebook Prophet

Facebook Prophet is popular, because its user-friendly and requires minimal configuration, while producing reliable results for business applications.

Unlike other time-series models, it can work well with messy data as well as irregular sampling which is common in the real world.

Preparing environment for Forecasting

Installation of Facebook Prophet

We can easily install Facebook Prophet by writing this command in the terminal

pip install prophet

Importing libraries

To start forecasting, we need to import the its library into our project.

import pandas as pd
from prophet import Prophet

Data to Forecast

Now our environment is ready, we need the data to forecast. We will start from forecasting the below data.

Forcasting using Facebook Prophet

Example 1: Forecasting Electric production

We have electricity production data in csv 1, we will predict the future electricity production using this data.

In the below code, we have imported Prophet and Pandas library. After that, we have stored csv data into a dataframe named df. We have them printed its last 5 rows.

import pandas as pd
from prophet import Prophet

#reading and storing csv data in a dataframe 
df = pd.read_csv('https://paradisecoder.com/wp-content/uploads/2024/12/Electric_Production-Sheet1.csv')

#printing dataframe last 5 values
print(df.tail())

Output of the above >>

        ds              y   
392	9/1/2017	98.6154
393	10/1/2017	93.6137
394	11/1/2017	97.3359
395	12/1/2017	114.7212
396	1/1/2018	129.4048

We have called Prophet model in variable m and then fittet our data in it which is a dataframe df.

#calling model
m = Prophet()

#fitting our data into the model
m.fit(df)

After that, we have made 10 additional data points in the future and stored it in future variable and printed its last 5 values using print function.

#making 10 data points in future for predicting
future = m.make_future_dataframe(periods=10)

#printing 5 values 
print(future.tail())

Output of the above >>

        ds
402	2018-01-07
403	2018-01-08
404	2018-01-09
405	2018-01-10
406	2018-01-11

Now, we have used this model to predict 10 values and store it in forecast variable.

Afterwards, we have printed last 5 values of the selected columns.

#forecasting data using model
forecast = m.predict(future)

#naming columns and printing it
print(forecast[['ds', 'yhat', 'yhat_lower', 'yhat_upper']].tail())

Output of the above >>

        ds              yhat            yhat_lower      yhat_upper
402	2018-01-07	119.983528	116.293831	123.725298
403	2018-01-08	121.357068	117.630464	124.935681
404	2018-01-09	122.864152	119.274042	126.459013
405	2018-01-10	124.450675	121.039400	128.114974
406	2018-01-11	126.060184	122.775125	129.886945