TASK : 01

LOAN ELIGIBILITY PREDICTION USING LOGISTIC REGRESSION

Objec ve:

 To understand the process of data preprocessing for machine learning.

 To implement feature engineering techniques.

 To build and evaluate a Logis c Regression model for loan eligibility predic on.

Tools/So ware Required:

 Python 3.x

 Pandas

 Scikit-learn (sklearn)

 Anaconda (recommended for environment management)

Dataset:

A simplified loan dataset is provided in the code. This dataset includes the following
features:

 income: Applicant's income.

 loan_amount: Requested loan amount.

 credit_score: Applicant's credit score.

 loan_approved: Loan approval status (1: Approved, 0: Not Approved).

Procedure:

 Data Preprocessing:

1. Load the Data:

o Create a Pandas DataFrame from the provided sample dataset.

o Print the original DataFrame to observe the raw data.

2. Handle Missing Values:

o Iden fy missing values in the 'income' column.

o Use SimpleImputer from scikit-learn to replace the missing values with

the mean of the 'income' column.
 o Print the DataFrame a er imputa on to verify the changes.

3. Scale Numerical Features:

o Iden fy the numerical features ('income', 'loan_amount', 'credit_score').

o Use StandardScaler from scikit-learn to scale these features to have zero

mean and unit variance.

o Print the DataFrame a er scaling to observe the transformed data.

 Feature Engineering:

1. Create a New Feature (Total Risk):

o Create a new feature named 'total_risk' by calcula ng the ra o of

'loan_amount' to 'credit_score'.

o Print the DataFrame a er adding the new feature to verify its crea on.

 Machine Learning Model Building and Evalua on:

1. Prepare Data for Model:

o Define the features (X) as the 'income', 'loan_amount', 'credit_score', and

'total_risk' columns.

o Define the target variable (y) as the 'loan_approved' column.

2. Split Data:

o Use train_test_split from scikit-learn to divide the data into training and
tes ng sets. Allocate 80% of the data for training and 20% for tes ng. Set
random_state=42 for reproducibility.

o Print the shapes of the training and tes ng sets to confirm the split.

3. Train the Model:

o Create a Logis cRegression model from scikit-learn.

o Train the model using the training data (X_train, y_train).

4. Make Predic ons:

o Use the trained model to make predic ons on the tes ng data (X_test).

o Print the predic ons (y_pred).

5. Evaluate the Model:

 o Use accuracy_score from scikit-learn to calculate the accuracy of the
model's predic ons by comparing them to the actual tes ng labels
(y_test).

o Print the calculated accuracy.

Program :

Python

import pandas as pd

from sklearn.model_selec on import train_test_split

from sklearn.linear_model import Logis cRegression

from [Link] import accuracy_score

from [Link] import StandardScaler

from [Link] import SimpleImputer

# 1. Sample Data

data = {

'income': [5000, 6000, None, 7000, 4500],

'loan_amount': [100, 150, 80, 200, 120],

'credit_score': [700, 750, 650, 800, 680],

'loan_approved': [1, 1, 0, 1, 0]

df = [Link](data)

print("Original DataFrame:\n", df)

# 2. Data Preprocessing

imputer = SimpleImputer(strategy='mean')

df['income'] = imputer.fit_transform(df[['income']])

print("\nDataFrame a er Impu ng Missing Income:\n", df)

scaler = StandardScaler()
numerical_cols = ['income', 'loan_amount', 'credit_score']

df[numerical_cols] = scaler.fit_transform(df[numerical_cols])

print("\nDataFrame a er Scaling Numerical Columns:\n", df)

# 3. Feature Engineering

df['total_risk'] = df['loan_amount'] / df['credit_score']

print("\nDataFrame a er Feature Engineering (total_risk):\n", df)

# 4. Model Building and Evalua on

X = df[['income', 'loan_amount', 'credit_score', 'total_risk']]

y = df['loan_approved']

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

print("\nTraining Data (X_train):\n", X_train)

print("\nTes ng Data (X_test):\n", X_test)

print("\nTraining Labels (y_train):\n", y_train)

print("\nTes ng Labels (y_test):\n", y_test)

model = Logis cRegression()

model.fit(X_train, y_train)

y_pred = [Link](X_test)

print("\nPredic ons (y_pred):\n", y_pred)

accuracy = accuracy_score(y_test, y_pred)

print("\nAccuracy:", accuracy)
Output :

Original DataFrame:

income loan_amount credit_score loan_approved

0 5000.0 100 700 1

1 6000.0 150 750 1

2 NaN 80 650 0

3 7000.0 200 800 1

4 4500.0 120 680 0

DataFrame a er Impu ng Missing Income:

income loan_amount credit_score loan_approved

0 5000.0 100 700 1

1 6000.0 150 750 1

2 5625.0 80 650 0

3 7000.0 200 800 1

4 4500.0 120 680 0

DataFrame a er Scaling Numerical Columns:

income loan_amount credit_score loan_approved

0 -0.727778 -0.715097 -0.301084 1

1 0.436667 0.476731 0.639803 1

2 0.000000 -1.191828 -1.241971 0

3 1.601112 1.668560 1.580691 1

4 -1.310001 -0.238366 -0.677439 0

DataFrame a er Feature Engineering (total_risk):

income loan_amount credit_score loan_approved total_risk

0 -0.727778 -0.715097 -0.301084 1 2.375075

1 0.436667 0.476731 0.639803 1 0.745121

2 0.000000 -1.191828 -1.241971 0 0.959626

3 1.601112 1.668560 1.580691 1 1.055589

4 -1.310001 -0.238366 -0.677439 0 0.351863

Training Data (X_train):

Income loan_amount credit_score total_risk

4 -1.310001 -0.238366 -0.677439 0.351863

2 0.000000 -1.191828 -1.241971 0.959626

0 -0.727778 -0.715097 -0.301084 2.375075

3 1.601112 1.668560 1.580691 1.055589

Tes ng Data (X_test):

income loan_amount credit_score total_risk

1 0.436667 0.476731 0.639803 0.745121

Training Labels (y_train):

4 0

2 0

0 1

3 1

Name: loan_approved, dtype: int64

Tes ng Labels (y_test):

1 1

Name: loan_approved, dtype: int64

Predic ons (y_pred):

[1]

Accuracy: 1.0

Results:

 Record the original DataFrame.

 Record the DataFrame a er each preprocessing step.

 Record the shapes of the training and tes ng sets.

 Record the predic ons made by the model.

 Record the accuracy of the model's predic ons.