Project Overview and Scope:

The goal of this project is to analyze loan dataset from a bank which contains information about it’s various borrowers. The main deliverable of the analysis is to find some attributes of borrowers who tend to be risky and shall be scrutinized properly before a loan is provided to them.

The dataset used in this project comes as a .json file and can be found here. I have cleaned the data using Python and then stored it in a csv format. The Python script can be found here. This script contains a detailed description of all the columns, which contain relevant information about various borrowers. I recommend anyone seeing this project, to first got through the Python Script to get an overall idea about the data and various columns contained in it.

Data Analysis Using Tableau:

Exploratory Analysis & Overall Summary

• More loans have been given at higher interest rates and number of borrowers tend to decrease with increase in utilization rates. This can be due to the fact that people with high utilization rates are viewed as risky borrowers by lending institutions.
• Majority of the borrowers come from the income group of (10k – 250k) per annum. It seems quite reasonable as people with incomes lower than 10k cannot afford to pay back at high interest rates. And people in higher income groups, above 250k, may not need to borrow money or do not need to borrow money for various purposes listed in the current dataset.
• Majority of the borrowers have Good or Very Good credit scores, and it is obvious because that is one of the main factors that determine whether any bank would lend a loan to a borrower or not.
• Borrowers who paid back fully had a Dti of 12.49, against a Dti of 13.2 for those who did not pay back in full.
• Those who do not pay back in full, tend to have higher values for Dti (Debt to Income) as well as higher Utilization Rates (%).

Some Insights based on Fico Category (Credit Score):

• 98% of the borrowers with fair credit score get loans at high interest rates. (above 12%)
• 346 out of 1341 borrowers having a fair credit score have not paid the debt in full, which is about 25%, which is a cause of concern.
• More borrowers with fair credit score tend to have a higher utilization rate, which in turn explains a low credit score for them. This holds for people who either pay or don’t pay their debts in full.
• The minimum interest rate charged to fair borrowers who did not pay back in full is 11.54%, whereas if we take all borrowers with fair credit score, then the minimum is 6%, which is also the global minimum.
• 51.8% of the borrowers with good or very good credit scores got a loan at low interest rate (below 12%).

Conclusion:

The data available does not present much insights into the attributes of a risky borrower, but overall the following are some points which can be looked for by any financial institution which wants to avoid risky borrowers:

• Borrower with Fair or below Credit Score
• Borrower with above average Dti (Debt to Income Ratio)
• Borrower with higher utilization rates

For further details, one can also refer the menu option, which gives the detailed records of borrowers under consideration on the dashboard.

However, to predict if a borrower is risky or not, will default or not, the best forward approach would be to build a binary classifier, like logistic regression or Random Forest Classifier, which can take in the data and predict the outcome based on new data that is used to train it.