• 

  Risky mortgages by State

• 

  Mortgage Purpose vs Risk

• 

  Risky mortgages vs Sellers

• 

  Risk Distribution for Neighborhoods

HackUTD2019

Problem Statement provided by Fannie Mae for Risk Analytics.

We provide our approach to find if the mortgage comes with Moderate, High or No Risk associated with it.

The solution predicts the possible 3 categories a loan may be associated with and the best results are:

High-Risk Loans

93% Precision and 96% Recall

Moderate Risk Loans

93% Precision and 93% Recall

No Risk Loans

91% Precision and 83% Recall

We wanted to use Area Under ROC and Area Under PRC as metrics but they are not suitable for multiclass classification problems and due to the limited time we used F1-Score which is a weighted score of PRECISION and RECALL.

For more details and Exploratory Data Analysis: https://utdallas.box.com/s/afh0zw1uyloffp1x2gre2ejbtb2tky7p

Some of the findings are:

1. More risky mortgages having properties in the state of Florida.
2. Most of the risky mortgages are associated with Bank of America.

Built With

• hadoop
• jupyter-notebook
• machine-learning
• pandas
• particle
• python
• scikit-learn
• spark2
• tableau