The data consists of information regarding 115,000 loan, including 81 variables The dataset can be found is provided by Udacity. In addition, I used US State map data from United States Census Bureau here, To use this data, I need to download geopandas and related libraries here.
I've found that Borrower APR and Credit Score Range Lower has negative moderate correlation. Borrower APR has moderate negative correlation with LoanOriginalAmount too. CreditScoreRangeLower and LoanOriginalAmount are moderately positively correlated.
Business Category is likely to have higher Credit Score. Loan status of Borrowers who have higher credit score are more likely to be completed. Also loans with lower BorrowerAPR are likely to be completed too which is also expected. CreditScoreRangeLower is generally lower for borrowers who has income verifiable.
The higher the income range the higher credit score. Borrowers with low income range likely to have higher Borrower APR.
States Maine (ME), North Dakota (ND) and Iowa (IA) has highest loan completion rates, whereas states Nevada (NV) and South Dakota (SD) has lowest. States South Dakota (SD), Alabama (AL) and Arkansas (AR) has highest Borrower APR whereas Maine (ME) and Iowa (IA) has lowest. States Wyooming (WY), South Carolina (SC), Virginia (VA), and Delaware (DE) has highest Credit Scores whereas North Dakota (ND) and Iowa (IA) has lowest. States Montana (MT), Wyooming (WY), and Idaho (ID), has highest debt to income ratio whereas California (CA), New York (NY), Virginia (VA) has lowest.
Category labeled as "Student Use" are more likely to have lower stated monthly income which may be related with lower credit scores.
Borrowers who has income not verifiable are more likely to apply for Business Category. The borrowers with income not verifiable has a lot higher proportion of Business Category than income verifiable.
Most of borrowers with income not verifiable are self employed.
Highest defaulted and chargedoff rates are in lower income ranges which is expected.
For the presentation, I started introducing three variables of interest: (Loan Status, Borrower APR and Credit Score Range Lower) then followed by two categorical variables, Income Range and Borrower State. Then I scatterplot two moderately correlated pairs with one of them is variable of interest.
Afterwards, I include categorical variables Loan Status, Is Borrower Home Owner Is Income Verifiable and Income Range into numerical variables of interest using heatmap (Credit Score Range Lower and Borrower APR)
I learned type of loan status from these resources: https://www.investopedia.com/terms/c/cancellation-of-debt.asp https://www.valuepenguin.com/loans/what-does-it-mean-to-default-on-a-loan https://www.investopedia.com/terms/c/chargeoff.asp
I used this stackoverflow source to convert value_counts to a dataframe: https://stackoverflow.com/questions/47136436/python-pandas-convert-value-counts-output-to-dataframe
To use value_counts in geopandas, I had to merge this state_count_df to a pandas dataframe.The source below was very helpful. https://towardsdatascience.com/lets-make-a-map-using-geopandas-pandas-and-matplotlib-to-make-a-chloropleth-map-dddc31c1983d Also I get help from this source: http://ramiro.org/notebook/geopandas-choropleth/
I've checked the below source to drop these states: https://stackoverflow.com/questions/13851535/how-to-delete-rows-from-a-pandas-dataframe-based-on-a-conditional-expression
To show states on the map, I need to get the representative point location in the map. Used this source https://stackoverflow.com/questions/38899190/geopandas-label-polygons/38902492
remove y axis: get help from: https://stackoverflow.com/questions/2176424/hiding-axis-text-in-matplotlib-plots
remove frames, get help from: https://stackoverflow.com/questions/14908576/how-to-remove-frame-from-matplotlib-pyplot-figure-vs-matplotlib-figure-frame
I've found from here a way to show proportions in y axis for bivariate exploration: mwaskom/seaborn#1027