Scribd
Upload
18 views6 pages

Decision Making Assignment

Uploaded by

ꜱʜᴜʙʜᴀᴍ ᴘᴀᴡᴀʀ
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
18 views6 pages

Decision Making Assignment

Uploaded by

ꜱʜᴜʙʜᴀᴍ ᴘᴀᴡᴀʀ
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 6

SUBJECT – DECISION MAKING IN

FINTECH
ASSIGNMENT 1
NAME – OMKAR RAUT
ROLL NO – 29
LOAN DELIQUENCY PROBLEM
Given:
Total number of customers bank has disbursed loan to between 2013 - 18 – 8,00,000
Loan delinquency rate – 10%

To find:
To identify customers whose account balance falls below a certain criteria
To identify 30,000 customers most likely to miss loan payments in next 3 months to send
timely reminders.
To study data of customers who missed payments (age, income, loan details).
To use customer behavior and characteristics to predict likelihood of missed payments.
To generate rules based on patterns and insights.
To evaluate whether reminders effectively reduced missed payments.
To explore how accurate and reliable are the predictions for real world applicable to real
world insights.

Step 1: Problem Definition


Among 800,000 personal loan borrowers, the bank is facing a 10% default rate, which
results in high fees and unhappy borrowers. Predicting which clients are most likely to skip
payments over the course of the following three months will help the bank send out timely
reminders.
•Important Goals Forecasting - Identifying 30,000 clients who are likely to skip payments
while staying within the contact center's reminder capacity is the aim.
•The company's objective is to decrease late payments and enhance client satisfaction by
avoiding delinquencies.
Goal Variable - Delinquency of Loans (Yes/No):
Yes (1): The client fails to make a payment within the following three months.
No (0): No payments are missed by the customer.

Step 2 : Data collection


(Data types)
1. Demographic Data
This information sheds light on the customers' individual traits, which may have an impact
on their propensity to miss payments.
•Age
•Gender
•Income
•Employment status
2. Loan Data
The customer's loan details give background information about their financial
responsibilities and the repayment plan.
•Loan amount
•Interest amount
•Monthly instalments
3.Transaction Data
The customer's financial behavior and liquidity, which are closely related to their capacity to
make payments, are shown in this data.
•Account Balance
•Transaction frequency
•Deposit and withdrawal patterns
4. Historical Data
The customer's past behavior, which is frequently indicative of future behaviors, is
contextualized by historical records.
•Payment history
•Credit score
•Loan history

Step 3: Data preparation


1)Checks for Data Quality:
•Missing Data: Use imputation or removal to deal with missing values.
*Outliers: Find and control outliers in important variables, such as account balance and
income.
•Consistency checker: Verify the consistency of the data from various sources by doing
consistency checks.
2)EDA, or exploratory data analysis:
•Distribution Analysis: Examine how important characteristics, such as age and income, are
distributed.
•Correlation: Determine the connections between the target variable and the
characteristics using correlation analysis.
•Visualization: To comprehend trends, patterns, and anomalies, use charts.
3)Feature Engineering:
•Derived Features: Develop fresh features like the balance to monthly instalment ratio.
•Scaling: To provide consistent model input, normalize numerical features.

Step 4 : Modeling
1)Analyzing Historical Data
Data Collection: Gather comprehensive historical data on customer characteristics,
transaction history, credit scores, and loan performance.
2) Defining Rules for the Model (Rule-Based Approach)
Objective: Create clear, actionable rules to flag high-risk customers.
Rule 1:
Condition:
Account Balance < 3 times the Monthly Installment
Number of Past Due Payments > 2 in the last 6 months
Credit Score < 600
Action: Flag the customer as High Risk.
Rule 2:
Condition:
Debt-to-Income Ratio > 0.5
Loan Amount > 5 times Monthly Income
Action: Flag the customer as High Risk.
Proactive Measure:
Send timely reminders to flagged customers two weeks before the due installment to
reduce the likelihood of missed payments.
3) Choosing a Statistical Model
Model Selection:
Decision Trees: For classifying customer behavior and identifying high-risk groups through
interpretability.
Logistic Regression: To predict the probability of loan delinquency based on defined
features.
Neural Networks: To capture complex, non-linear patterns and relationships between
customer data and delinquency outcomes.
4) Training the Model
Data Splitting:
Split the dataset into training (80%) and validation (20%) sets to ensure robust model
evaluation.
Feature Matrix:
Define the feature matrix with key variables such as Account Balance, Credit Score, Debt-to-
Income Ratio, Loan Amount, and Past Due Payments.
Model Training:
Train the model on the training dataset.
Use the validation set to evaluate performance and fine-tune hyperparameters.
Model Evaluation:
Evaluate model performance using metrics like Accuracy, Precision, Recall, and F1 Score.
Implement cross-validation to ensure the model generalizes well to unseen data.
Final Testing:
Apply the trained model to the test dataset (remaining 50%) to identify potential loan
delinquents.

Step 5: Model Evaluation


It helps in evaluating performance of the chosen model and analyse any risen error.
 Sources of error in loan delinquency big data model –
Missing/incomplete data
Overreliance on a single metric
Selecting complex models
Errors are predicted and analysed via regression, decision trees and confusion matrix.
A model must achieve low bias and variance for high accuracy.

Step 6- Experimentation
To understand the quantify actual impact of models in the real world.
i. Defining relevant success metrics (repayment rate)
ii. Dividing population into two random groups (A/B Testing) divide the dataset containing
information on loan holders into two groups: one receives reminders (control group) and
one does not (treatment group)
iii. Execute experiment (in real world i.e. banks, private loan companies)
iv. Measure success metrics : Track delinquency rates, false positive/negative rates, other
KPIs for both groups and compare the delinquency rate in the treatment group with the
control group.

In concluding, I think Big data model will give the accurate result as all possible
consequences and situations have taken into consideration to detect and prevent loan
delinquency.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy