This post is authored by Bharath Sankaranarayan, Principal Program Manager at Microsoft.
I am pleased to announce the Loan Credit Risk Solution how-to on
SQL
and
HDI
. Loan Credit Risk is a very common problem that lending institutions are challenged in assessing which loans are likely at risk.
Loan Credit Risk solution how-to, demonstrates how you can apply machine learning to manage loan credit risk – deciding who to issue loans based on prediction of defaulting risk (at the time of loan application).
For loans already issued, you can check our solution template Loan ChargeOff Prediction on
SQL Server
and
HDI
.
The solution is built using Microsoft R Server 9.1 (April 2017) on both SQL Server 2016 and HDI on Spark (Microsoft R server 9.1 released in June 2017). For detailed information and the features and capabilities you can check out the blog -
Microsoft R Server 9.1
The solution how-to was developed for small personal loan financial institution using the borrower's historical financial data coupled with the information while requesting for a loan. The solution how-to was built to cater two personas, the
business user
in this case a loan officer and the
data scientist
who is interested in how the data was used to build the models used for the calculating the risk. The goal is to help the lending institution reduce risk on their investment.
To the loan officer, to decide how much risk the institution can take based on the outstanding loans and what are the limits, an interactive dashboard that provides visibility to the test data that was used and a tab that shows the new potential loans with a cutoff value (the risk level that the institution is willing to take). These two provide the loan officer the ability to make the decision on the loan. The output or the predictions (scores) are grouped in percentiles (the higher the percentile, higher the risk).
You can try this interactively via the Try it now button for either the
SQL
or the
HDI
solution and here is a screen shot of the
power bi
dashboard.
|
|
Now to the data scientist who is interested in how these are developed we have provided detailed steps to follow through. The solution walks through the
Data Science Process
and shows how to train and score the models in both
SQL Server
and
HDI on Spark Cluster
. These are provided in an easy to follow steps for both platforms. The SQL Server implementation leverages the
Data Science VM for Windows
which provides all the tools you need to try the solution in a single box.
The completed solutions are deployed to SQL Server 2016 by embedding calls to R in stored procedures. These solutions can then be further automated with SQL Server Integration Services and SQL Server agent.
You can also use PowerShell scripts or Jupyter Notebooks, in addition to using IDEs such as R Tools for Visual Studio. We have made the entire code that powers this solution free to use and modify as well.
When you deploy the
SQL Server
based solution, you will get to experience Jupyter Notebooks in additional to the R IDE.
To try this out please visit
Loan Credit Risk Solution how-to on SQL
or
Loan Credit Risk on HDI
using Spark and send us your
feedback
to let us know how we are doing. You can also check out our other solutions
here
.
Bharath