Lending is one of the core businesses in any Banking/Financial Industry. Banks lends money in various forms like Credit Card, Retail Loans (Housing Loan, Car Loan, Personal Loan etc..), Corporate

Loans,  Trade Finance etc..

Machine Learning project developed will be used to predict the Credit Risk before

sanctioning loan and during life cycle of the loan.

Data pipeline Data pipeline build to

process the customer application details through sequence of ML process and

evaluate customer credit worthiness, credit behavior in the past, credit

details shared among the financial institution, etc..

Data Pipeline starts from Data Ingestion, Data Validation, Data preprocessing (Data cleaning, Data engineering) Data transformation, Model Training and Validation and Prediction

Data Ingestion - Data collected through various systems from various countries for many years are stored in Banks Data warehouse system. Since the Information/data evolved over the period, data was not in

standard format Data Cleansing was a major challenge.

ML Training/ML Analysis Applied various Machine Learning methods like Linear Regression, decision Tree, Random Forest, SVM, AdaBoost and Deep Learning Technique to train the ML Model and analyzed the

results.

Customer using various banking service to transact on a daily basis like online baking, mobile

banking, online shopping, credit/debit card, etc,. Problems, issues, concerns

and suggestions faced by customers are reported to bank through online, phone

and hand written modes.   These are valuable information for the bank to address the customer compliant on time and to improve the business strategy.

Content of the compliant will have the structured and unstructured information like text and

image. Historical Customer complaints are used for training the machine

learning algorithms to classify the issues. During training used BERT Question-Answering, Text

Classification models.

Customer using various banking service to transact on a daily basis like online baking, mobile

banking, online shopping, credit/debit card, etc,. Problems, issues, concerns

and suggestions faced by customers are reported to bank through online, phone

and hand written modes.   These are valuable information for the bank to address the customer compliant on time and to improve the business strategy.

Content of the compliant will have the structured and unstructured information like text and

image. Historical Customer complaints are used for training the machine

learning algorithms to classify the issues. During training used BERT Question-Answering, Text

Classification models.

It covers the entire lifecycle of Loans for Retail and Corporate customer. System has the capability to parameterise various Loan products like Housing, Auto, and Deposit Linked with feature of handling various types of interest calculation, Fixed/Prime Interest Rate; Index based floating Interest Rate, Deposit Linked Interest Rates, etc.   Payment schedule generation and accrual are one of the unique selling points of the system. System has the flexibility to reschedule or restructure the loan during the lifecycle. It supports multi-Lingual, multi-currency and multi branches.  Loan Top-up, Rollover and Component Rollover are key additional feature of the system.