Chat with us, powered by LiveChat

Credit Calculator

The main point of the project is to calculate the credit interest rate for a user. The purpose of the project was to find out the best interest rate for a loan to satisfy both investors and borrowers.

  • Service: Web App
  • Industry: Finance


  • Python 3.5
  • Django
  • Scikit-Learn
  • Pandas
  • NumPy
  • Celery
  • Jenkins
  • SonarQube
image description


This is done with Machine Learning which takes the available data about a user and predicts his ability to pay the debt back. Based on the classification given by the machine learning system, the simulations are run for thousands of similar users for each combination of term and amount to define how many of them would be able to pay in time, how many would repay earlier (thus not paying additional interest and servicing fees), how many would be delinquent, and how many won’t be able to pay at all.

How It Works

Based on the simulations, the most appropriate interest rate is chosen: high enough to attract investors and low enough to attract borrowers. The options which the user is unlikely to handle are filtered out, and only the probable ones are suggested to the user so he could choose the one he finds the most useful.
The system also calculates all the kinds of payments which are included in the loan (principal, interest, servicing fee, etc.) for the user for every time period to see what he pays for. The simulations are run for thousand of users to estimate the probability of receiving a certain amount of money from scheduled payments for each time period.
The internal modules are connected through the API which allows the communication between the modules and internal services. The command line interface helps with experimenting with different machine learning algorithms and ways of estimating the user’s credit grade.

image description

Additional Details

Our developers also provided the presentations of the calculations and writing the documentations, describing the process. The intermediate values of calculations were presented in different formats (json, html, xlsx) for evaluating models for correctness.


"They are very solution oriented and figured out how to make things work for everyone as opposed to pointing fingers and deflecting."

Oliver Centner CEO at UnoAPP.