Checkout our work as a Streamlit WebApp.

This is a Machine Learning Centric project. In this project, we have trained different ML models on four datasets and deeply analyzed all the data from the AA to come up with 5 major features that might help stakeholders(banks, business, etc.) as well as the customers. The 5 features we are offering are briefed below:

Customer Segmentation: It allows banks, businesses and relevant stakeholders to tailor their products and services to each customer segment's specific needs and preferences based on several data points collected through AA. The data points include: age, income, average account balance, credit history, investment history and transaction data. This feature can make the stakeholder’s offerings more appealing to target customers, which can help the bank to acquire new customers and retain existing ones. Use Case: taking BoB as an example, if a customer is a high-spending user, then we can recommend them Baroda Salary Premium Account. If the person is a senior citizen then Baroda Senior Citizen Privilege Scheme

Loan Defaulter Detection: This feature helps banks in detecting potential defaulters by going through their financial history. Currently, the process is done manually by humans, but now with the help of AA all the Account histories are available in one place, we apply our model to predict if the customer is eligible for a loan or not. This process is more automatic and accurate with a lot more data such as transfer history, equity data, debt-to-income ratio, loan amount, length of credit history, etc.

Cashflow Analysis/ Purchasing Power: This is a feature aimed at customers who are aiming to achieve their financial goals. We analyze their transaction history and show them the cashflow chart such as: Food, entertainment, grocery, healthcare, EMI, etc. This will help them to cut down costs in certain areas. This analysis also helps us in determining in what we call Purchasing Power. Purchasing Power is a score given to people who don’t have a credit history based on their financial data. Right now there are many Buy now Pay Later apps such as Slice, Jupiter, One card, Simpl, etc. that give out credit with almost no basis. Instead, Purchasing Power Score will help them determine an appropriate credit amount to be given to the customers.

Fraud detection: Trained on a flagged fraud transaction dataset, our model will be able to detect fraudulent transactions in the history of the user, this can help in marking them as a person of interest as well as denying financial services to that particular person. This can help in decreasing in the number of bounces, denied requests as well as bad transactions.