This application gives the bank an advantage to contact customer that is really a good fit and at the right time and offer the loan that the customer really needs and in high possibility will take.
What it does
Help bank find potential clients for a new loan by using machine learning KNN algorithm to find k neighbors that their transactions are closest to the customer we examine and check if they took the loan, the percentage of the loan probability is calculated by the percentage of the k neighbors that took a loan. The initial data is taken from FFDC (due to poor amount of the data we also added some data manually to the DB.). This application gives the bank an advantage to contact customer that is really a good fit and at the right time.
How we built it
There is a process that get the data from FFDC and parses it to the DB. After that there is a python script that runs the machine learning on this data and builds analyzed data and stores it in another table in DB. Then there is a UI that allows user to see the customers, their data and contact details and the possibility that they will take the loan. If you click on the customer row, you will get the detailed information about the customer, account and transactions. It is also possible to run a script that sends email/ sms/ watsapp to customers that have loan possibility higher than 60 percent(for example).
Challenges we ran into
To make something working as we wanted in a very short time
Accomplishments that we're proud of
Machine learning, cloud ready, a great UI, future ideas that can be implemented for this system.
What we learned
New technologies, to build such system from scratch in a very short time.
What's next for The magic of the loan prediction
Cloud, automatic messages to customers, additional analysis on the data, additional topics besides loans....