Sorry for the mute video, my mic just gave up on me
Wanted to try out new things in finance.
What it does
Predicts whether a customer will default on their Credit Card Bills.
How we built it
(1) First we got the data set from kaggle. The data set values are displayed in the first dataframe chart on the web app (2) Then I visualized the relations between the different features to find out which feature values affect the other ones the most. (3) Then we had to scale the values and split it into test and train (4) We trained 3 different models : Logistic Regression, Support Vector Machine and MLP classifier (5) I saved the models using pickle for further use (6) The three excel files contain dataset about 5,5 and 10 customers respectively. In real world scenarios, the analyst working for the Financial Institutions would just upload one on the website. (7) We then feed this excel file to our models which predicts whether the customer will default or not. (8) We can choose between the algorithms. (9) Finally we visualize between the relation between the credit card bill amounts and the amounts paid subsequently
Challenges we ran into
The Web interaction part was a bit tricky cause we are all backend devs but god damn it streamlit is just too good. Models were not working on newer versions of different libraries.
Accomplishments that we're proud of
Made it work before the deadline , Phew!!.
What we learned
Prediction Models, Correlation between the features, Visualization of Data.
What's next for Credit Card Default Prediction
To add more features like a User Credit Score Calculation and Personalization.