Predicting Loan Approval using Classification Models

Objective

The goal of this project is to construct a machine learning model that predicts Loan Approval. We'll explore five different classificaton models:

• K-Nearest Neighbor (KNN)
• Support Vector Machine (SVM)
• Decision Tree
• Random Forest
• XGBClassifier

Each model will be carefully evaluated using cross-validation and performance metrics such as F1 score. After identifying the best base model, we'll fine-tune its parameters to improve its accuracy too. Finally, we'll compare the tuned model with the base one to choose the best option for predicting loan approval accurately and reliably.

Dataset Information

This project utilizes a raw dataset sourced from kaggle.com titled "Loan Approval Dataset" uploaded by Rohit Sharma. The dataset comprises information on various attributes of loan applicants, including demographic details, financial status, employment history, and ownership status. The dataset includes both numerical and categorical features, making it suitable for diverse analytical approaches. For access to the dataset source, you can click here!

Dataset Description Table

Column Name	Description
ID	Unique identifier for each loan applicant.
Income	The income level of the applicant.
Age	Age of the applicant.
Experience	Years of professional experience.
Married/Single	Marital status of the applicant.
House_Ownership	Indicates whether the applicant owns or rents a house.
Car_Ownership	Indicates whether the applicant owns or rents a house.
Profession	Occupation or profession of the applicant.
City	City of residence of the applicant.
State	State of residence of the applicant.
Current_Job_Yrs	Duration of employment in the current job.
Current_House_Yrs	Duration of residence in the current house.
Risk_Flag	Binary indicator of loan risk, where 1 represent a flagged risky applicant and 10 represents a non-risky applicant.

---

Sample Dataset

---

Raw Dataset

Id	Income	Age	Experience	Married/Single	House_Ownership	Car_Ownership	Profession	CITY	STATE	CURRENT_JOB_YRS	CURRENT_HOUSE_YRS	Risk_Flag
1	1303834	23	3	single	rented	no	Mechanical_engineer	Rewa	Madhya_Pradesh	3	13	0
2	7574516	40	10	single	rented	no	Software_Developer	Parbhani	Maharashtra	9	13	0
3	3991815	66	4	married	rented	no	Technical_writer	Alappuzha	Kerala	4	10	0
4	6256451	41	2	single	rented	yes	Software_Developer	Bhubaneswar	Odisha	2	12	1
5	5768871	47	11	single	rented	no	Civil_servant	Tiruchirappalli[10]	Tamil_Nadu	3	14	1

Final Dataset for Modeling

Married/Single	House_Ownership	Car_Ownership	CITY	Age	Experience	CURRENT_JOB_YRS	CURRENT_HOUSE_YRS	Risk_feature
1	2	1	45	60	10	5	11	0
1	2	0	280	66	18	12	10	0
1	2	0	187	59	14	14	14	0
1	2	0	303	32	3	3	11	0
1	2	0	162	21	18	8	12	0

Model Performance

Five Model Performance Comparison

Metric	KNN	SVC	Decision Tree	Random Forest	XGBoost
Train Score	0.801	0.571	0.929	0.921	0.805
Test Score	0.7632	0.502	0.82	0.835	0.748
Precision Train	0.802	0.561	0.898	0.893	0.794
Precision Test	0.314	0.146	0.405	0.412	0.298
Recall Train	0.799	0.648	0.967	0.963	0.825
Recall Test	0.787	0.631	0.858	0.839	0.779
F1 Score Train	0.800	0.601	0.931	0.927	0.809
F1 Score Test	0.450	0.237	0.551	0.553	0.432
Cross Validation F1 Mean	0.763	0.594	0.835	0.830	0.763

Random Forest vs Tuned Random Forest Performance

Model Name	Random Forest	Tuned Random Forest
Train Score	0.921	0.571
Test Score	0.835	0.502
Precision Train	0.893	0.561
Precision Test	0.412	0.146
Recall Train	0.963	0.648
Recall Test	0.839	0.631
F1 Score Train	0.927	0.601
F1 Score Test	0.553	0.237
CV F1 Mean	0.830	0.594