Building a Logistic Regression

  

Building a Logistic Regression

---

Create a logistic regression based on the bank data provided.

The data is based on the marketing campaign efforts of a Portuguese banking institution. The classification goal is to predict if the client will subscribe a term deposit (variable y).

Note that the first column of the dataset is the index.

Import the relevant libraries

---

import pandas as pd

import numpy as np

import statsmodels.api as sm

import matplotlib.pyplot as plt

import seaborn as sns

sns.set()

 

# this part not be needed after the latests updates of the library

from scipy import stats

stats.chisqprob = lambda chisq, df: stats.chi2.sf(chisq, df)

 

 

 Load the ‘Example_bank_data.csv’ dataset.

from google.colab import files

uploaded = files.upload()

raw_data = pd.read_csv('Example_bank_data.csv')

raw_data

 

We want to know whether the bank marketing strategy was successful, so we need to transform the outcome variable into 0s and 1s in order to perform a logistic regression.

# We make sure to create a copy of the data before we start altering itNote that we don't change the original data we loaded.

data = raw_data.copy()

 

# Removes the index column that came with the data

data = data.drop(['Unnamed: 0'], axis = 1)

 

# We use the map function to change any 'yes' values to 1 and 'no' values to 0. 

data['y'] = data['y'].map({'yes':1, 'no':0})

data

 

# Check the descriptive statistics

data.describe()

Declare the dependent and independent variables

 

y = data['y']

x1 = data['duration']

 

Simple Logistic Regression

x = sm.add_constant(x1)

reg_log = sm.Logit(y,x)

results_log = reg_log.fit()

 

# Get the regression summary

results_log.summary()

 

 

# Create a scatter plot of x1 (Duration, no constant) and y (Subscribed)

plt.scatter(x1,y,color = 'C0')

 

# Don't forget to label your axes!

plt.xlabel('Duration', fontsize = 20)

plt.ylabel('Subscription', fontsize = 20)

plt.show()

 

 

np.set_printoptions(formatter={'float': lambda x: "{0:0.2f}".format(x)})

#np.set_printoptions(formatter=None)

results_log.predict()

 

np.array(data['y'])

results_log.pred_table()

 

cm_df = pd.DataFrame(results_log.pred_table())

cm_df.columns = ['Predicted 0','Predicted 1']

cm_df = cm_df.rename(index={0: 'Actual 0',1:'Actual 1'})

cm_df

 

cm = np.array(cm_df)

accuracy_train = (cm[0,0]+cm[1,1])/cm.sum()

accuracy_train