AI Banking Suite

Azure Developer League Hackathon

---

Objective

AI Banking Suite is a set of Azure AI services that makes the banks smarter and allow them to operate efficiently.

Implementation

Following is the invincible set of services:

• Smart Loan Approval Application: Using Azure Responsible AI, we are going to train a model, which will help in conducting the initial screening of the Loan Application Form, and shortlist the eligible ones for Loan Approval.
• 24x7 Smart Banking Bot: Trained on Azure Bot services, which will interact with the customers 24x7 and help in resolving all the initial queries about the services, site's navigation, customer care, etc.
• Credit Card Fraud Detection System: Trained on Azure ML Studio's supervised training model, from the previous transaction, flagged with fraud transaction, it will help in model training and prevent the transactions from fraud credit cards.
• Predictive AI: Monitors the bank growth constantly and predicts its future positions in the market. This will help the banks in tracking their current progress and modify their current methodologies if need be.
• Permanent Account Closure: Designed for the bank's clients, we have seen account opening facilities available so that customers can easily open a bank account digitally, but no bank offers permanent account closing facility. We provide this facility in our project ensuring secure conduct through facial recognition!

Detailed Working

1. Smart Loan Approval Application: This is the Machine Learning Model trained on VotingEnsemble Algorithm using Azure Automated ML. Loan Applicant will be given an online form to fill. On the basis of data filled by the customer, the model will recommend to the banks, whether or not, the Applicant should be given the loan or not. Details of the customers, alongwith the response will be saved to a separate Database, accessed by the Bank. Note: Services are being stopped, due to incurring charges. Placeholders are used in place. CGI-Scripts are mentioned in this repo

LoanApproval.mp4

2. 24x7 Smart Banking Bot: This is the FAQ virtual agent, that most of the banks used, for their technical assistance. This bot is trained on Microsoft Power Virtual Agents and deployed in the web app. It is available for customers 24x7 and help in resolving all the initial queries. Note: Services are being stopped, due to incurring charges. Placeholders are used in place.

bot.mp4

3. Credit Card Fraud Detection System: This is a solution for financial sector, but not bank specific. Many e-commerce sites are facing the issues of fraudulent transactions. So, we have trained a regression model on the credit card transaction demo dataset, and then this model will help in identify the corruot transaction. This model is trained on more false positive data, So it may not be accurate as much. But our team is working on this model, to make it realtime streaming data model rather than batch streaming. Note: Services are being stopped, due to incurring charges. Placeholders are used in place. CGI-Scripts are mentioned in this repo

creditModel.mp4

4. Permanent Account Closure via. Online Application: We offer an online application, for permanent account closure request trough our online portal. User have to provide the information in the fields given. To add an extra layer of security, we have added the power of AI to this application. We have enabled paperless signature verification as well as Facial Verification as the consecutive steps before one can submit the form for approval. Note: Models and frontend are ready, attached in this repo. Connecting frontend to the models is yet to be done.

closureform.mp4

5. Bank Growth Prediction: It is an ML model trained on a Bank's economic status and banking and systemic crisis data where inflation, currency crisis and bank crisis of 13 African countries between 1860 to 2014 is given to us. This model is trained on Azure Machine Learning Studio, then it is consumed on Power BI, which creates a powerful report and further dashboards, which we have published in the web app. Note: To publish Power BI dashboards in web app, we need Power BI Professional. That is why, after creating the service and the promo video, we have stopped the service to save the charges

predictionBI.mp4

Contribution:

This project is designed by Team Connoisseurs:

1. Abhinav Sharma
2. Aditi Gupta
3. Aditya Khandelwal
4. Akshat Soni