Scribd
Upload
5 views12 pages

Risk Segmentation Using Machine Learning

The document discusses the implementation of a machine learning model for risk segmentation in underwriting, aiming to enhance efficiency and accuracy in financial decision-making. It outlines the challenges faced in traditional underwriting processes and presents a proposed solution that integrates machine learning to reduce manual efforts and processing times. Key results include a significant reduction in application turnaround time and improvements in operational efficiency.

Uploaded by

panchariyapiyush2013
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
5 views12 pages

Risk Segmentation Using Machine Learning

The document discusses the implementation of a machine learning model for risk segmentation in underwriting, aiming to enhance efficiency and accuracy in financial decision-making. It outlines the challenges faced in traditional underwriting processes and presents a proposed solution that integrates machine learning to reduce manual efforts and processing times. Key results include a significant reduction in application turnaround time and improvements in operational efficiency.

Uploaded by

panchariyapiyush2013
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 12

Risk

Segmentation
using Machine
Learning

BY: P I Y U S H PA N C H A R I YA
Flow of Presentation

 Introduction

 Application Flow and Underwriter Assessment

 Challenges

 Proposed Solution

 Model

 Results and Impact

 Appendix
Introduction

 Bridging the gap between traditional underwriting and data-driven risk


assessment, making financial decision-making more efficient, accurate,
and scalable.
Application Flow and Underwriter
Assessment
STP Apps
Risk Team

Application Bank
Risk + Underwriter +
Call center

Underwriters

Underwriter
Challenges
Difficulty in
Higher Higher handling
Processing Operational large
Time Cost volume

Risk of Slower to
Human respond to
Error market
changes
Proposed Solution

STP Apps
Risk Team

Application Bank
Risk + Underwriter +
Call center

Machine
Learning
Model Underwriter
Risk Segmentation Model

Applicatio High
n Data

EDA and Model


Transactio Master Data Feature Selection Medium
n Data
Engineering and Result

Bereau Low / No
Data Risk

Input Machine Learning Output


Model
Master Dataset

• Customer Information: Age, Gender, Income, Distribution of Dependent Variable


567439

Address, Account Creation Date, Etc.


• Transaction Data: Transaction ID, Timestamp, 381664

Amount, Payment Method, Location, Etc.


• Beauro Data: Products, Defaults, Banks, Loans,
Cards, Credit Utilization, Credit enquiries, Etc.
50897

• Shape: ~1.2 Million x 120 features


High Medium Low
EDA

• Treated missing values.

• Outliers Identified and Treated.

• Treated Categorical Features.

• Performed Univariate and Bivariate Analysis to identify trends and patterns.

• Identified that high value transactions late night transactions are more risky transactions.

• Created Features: Transaction Velocity, Product Diversity, Peak Transaction Hours, Count
of defaults, consecutive defaults, etc.
• Standardized the data and Performed PCA.
Model Selection

 Multiclass Logistic Regression (F1 Score: 0.89, Recall 0.91)


 One vs Rest Multiclass Classification

 XGBoost Classifier (F1 Score: 0.92, Recall 0.95)


Results and Business Impact

 Key Improvements Achieved:


• ✅ Reduced Manual Efforts by 30% (2.4 hrs/underwriter/day)

• ⏳ Application TAT reduced by 45 minutes, improving customer satisfaction

• 📈 Increased Overall Operational Efficiency (OOE)


THANK YOU

You might also like