In [1]:
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from fbprophet import Prophet
%matplotlib inline
In [2]:
df = pd.read_csv('avocado.csv')
df.head(20)
Out[2]:
In [3]:
df.info()
In [4]:
#Renaming numerical column names. Saving changes to original dataframe.
df.rename(index=str, columns={"4046": "Small Hass", "4225": "Large Hass", "4770": "Extra-Large Hass"}, inplace=True)
In [5]:
#Dropping Unnamed: 0 column
df.drop(['Unnamed: 0'], axis=1, inplace=True)
In [6]:
#Making data column readable by Pandas
df['Date'] = pd.to_datetime(df['Date'])
In [7]:
df.head()
Out[7]:
Analyzing categorical data to fully see the range of this set
In [8]:
df['region'].value_counts()
Out[8]:
In [9]:
df['region'].nunique()
Out[9]:
In [10]:
df['year'].value_counts()
Out[10]:
In [11]:
df['type'].value_counts()
Out[11]:
Here we see that there are two different types of avocadoes: Conventional and Organic. Lets separate these two types from eachother.
In [12]:
conventional_frame = df[df.type == 'conventional']
organic_frame = df[df.type == 'organic']
In [13]:
organic_frame.info()
In [14]:
organic_frame['AveragePrice'].mean()
Out[14]:
In [15]:
conventional_frame['AveragePrice'].mean()
Out[15]:
In [16]:
organic_frame[['AveragePrice','Small Hass']]
Out[16]:
Finding how many regions are recorded for each type of Avocado as well as how many entries per region
In [17]:
regions_organic = organic_frame.groupby(organic_frame.region)
print("Total regions for Organic avocado:", len(regions_organic))
print("-------------")
for name, group in regions_organic:
print(name, " : ", len(group))
In [18]:
regions_conventional = conventional_frame.groupby(conventional_frame.region)
print("Total regions for Conventional avocado:", len(regions_conventional))
print("-------------")
for name, group in regions_conventional:
print(name, " : ", len(group))
Now we can make predictions based on a specific region we choose from either conventional or organic avocados. Lets start with organic and choose the "TotalUS" region.
In [19]:
date_price = regions_organic.get_group("TotalUS")[['Date', 'AveragePrice']].reset_index(drop=True)
In [20]:
#fig, ax = plt.subplots(figsize=(15,10))
date_price.plot(x='Date', y='AveragePrice', kind="line",figsize=(15,10))
plt.savefig('line_organic_avgp.png',bbox_inches='tight')
In [21]:
#Renaming the columns so they work with the fbprophet library
date_price = date_price.rename(columns={'Date':'ds', 'AveragePrice':'y'})
In [22]:
#Creating & fitting a model. All of the code here is further explained on prophet quick start page https://facebook.github.io/prophet/docs/quick_start.html
m = Prophet()
m.fit(date_price)
Out[22]:
In [23]:
# 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.
future = m.make_future_dataframe(periods=365)
In [24]:
#This shows the last dates that will be forecasted up to
future.tail()
Out[24]:
In [25]:
# 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.
forecast = m.predict(future)
forecast[['ds', 'yhat', 'yhat_lower', 'yhat_upper']].tail()
Out[25]:
In [26]:
# You can plot the forecast by calling the Prophet.plot method and passing in your forecast dataframe.
fig1 = m.plot(forecast)
plt.savefig('forecast_organic.png',bbox_inches='tight')
Here we can see how the individual components of the model affect the predictions
In [27]:
# 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.
fig2 = m.plot_components(forecast)
plt.savefig('forecast_organic_components.png',bbox_inches='tight')
Now lets instead analyze conventional avocados in the TotalUS region.
In [28]:
date_price = regions_conventional.get_group("TotalUS")[['Date', 'AveragePrice']].reset_index(drop=True)
In [29]:
date_price.plot(x='Date', y='AveragePrice', kind="line",figsize=(15,10))
plt.savefig('line_conventional_avgp.png',bbox_inches='tight')
In [30]:
date_price = date_price.rename(columns={'Date':'ds', 'AveragePrice':'y'})
In [31]:
#Creating & fitting a model. All of the code here is further explained on prophet quick start page https://facebook.github.io/prophet/docs/quick_start.html
m = Prophet()
m.fit(date_price)
Out[31]:
In [32]:
# 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.
future = m.make_future_dataframe(periods=365)
In [33]:
#This shows the last dates that will be forecasted up to
future.tail()
Out[33]:
In [34]:
# 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.
forecast = m.predict(future)
forecast[['ds', 'yhat', 'yhat_lower', 'yhat_upper']].tail()
Out[34]:
In [35]:
# You can plot the forecast by calling the Prophet.plot method and passing in your forecast dataframe.
fig1 = m.plot(forecast)
plt.savefig('forecast_conventional.png',bbox_inches='tight')
In [36]:
# 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.
fig2 = m.plot_components(forecast)
plt.savefig('forecast_conventional_components.png',bbox_inches='tight')
