SOCIB - Social Credit Information Bureau
Know your real credit score based on your Social Media Content with your consent.
Created on 20th October 2019
•
SOCIB - Social Credit Information Bureau
Know your real credit score based on your Social Media Content with your consent.
The problem SOCIB - Social Credit Information Bureau solves
Today a person having a monthly salary of say Rs.10,000 can have a higher CIBIL or credit score than a person having Rs.2,00,000. But, this doesn't speak about the real REPAYMENT capacity of a person.
So, CIBIL says 1st user is better than 2nd. But, by metrics, the 1st user is the one who has the highest chances of defaulting the First big loan and eventually leading to NON-PERFORMING ASSETS (NPA's) for FINANCE COMPANIES/BANKS
Social media really speaks of who you are-
- Your photos of car, dress, etc., - we do Convolutional Neural Network
- Your location data - we do correlation of how much you spend on trips and leisure
- Your post texts - we do Entity Extraction, Correlation and Entity processing(further pipeline)
Why would you give your data? (By YOUR OWN CONSENT)-
- Say you are planning to buy a house worth Rs.2Cr. Your CIBIL score is 750 and the bank says then can lend you Rs.1.25Cr, that's what your CIBIL says. But, since you know you can save more and pay back more than what your Bank thinks, we can help you here! :) We analyse your real capacity and help you get that buffer of Rs.75L.
- For Banks or Finance Companies, they need this to identify the earlier mentioned case of - Low Income + High CIBIL Score. This has to be a red flag and the Loan amount has to reduce based on how much he can payback.
NOTE: VIDEO IS AT GITHUB, please watch it.
Challenges we ran into
We have code written in NODEJS(webapp), PYTHON(NLP and CNN(, Java Spring(Core Backend). Most of our time went in organizing and getting the workflow to work. Also, our stuff is hosted in AWS. Lot of time went in setting up stuff on local laptop to develop and later resetup stuff on AWS.
It was a pretty huge project, we haven't slept one bit! :) And we are happy with the result.
Technologies used