Machine Learning Driven Models for Business Efficiencies

Machine learning buzz has been around for a few years and the technology is gaining broader adoption. See below the related Gartner Hype Cycle chart. One key driver for mainstream adoption outside of large corporations (who have the budgets to hire data scientists, machine learning engineers, etc.) is the maturation of the toolset provided by tech providers such as Google Cloud Platform. GCP’s tools provide the ability for machine learning applications to be built, managed, and maintained by mainstream technology and app dev teams without requiring an army of specialized skill sets. In this initial blog post, I cover some examples of the benefits of incorporating machine learning into the business toolkit for improving and optimizing operations.

As a quick sidebar and strolling down memory lane (hopefully without dating myself!) my first exposure to the field was during my time at Oklahoma State University where I was part of a team that developed a model for predicting box office revenues for Hollywood films based on parameters such as the month of release, cast, director, movie genre, box office history of cast and director, and etc. We purchased the dataset from a distributor in Los Angeles who [wait for it…..] shipped it to us via snail mail in floppy disks. I can imagine some readers googling “what is a floppy disk” We’ve come a long way!

What Is Machine Learning?

Terminology and definitions can be complicated so let’s start with a simple definition from Aurélien Géron – “Machine learning is the science (and art) of programming computers so they can learn from data”

ML is a subset of the larger field of artificial intelligence (AI) that “focuses on teaching computers how to learn without the need to be programmed for specific tasks,” per Sujit Pal and Antonio Gulli. “In fact, the key idea behind ML is that it is possible to create algorithms that learn from and make predictions on data.”

Examples of ML include the spam filter that flags messages in your Gmail, the recommendation engine Netflix uses to suggest content you might like, and the self-driving cars being developed by Google and other companies.

Figure 1 - Machine Learning Driven Models for Business Efficiencies SpringML

Underwriting/Risk Evaluation

Within the mortgage industry, automated underwriting has been around since at least the mid-90s. GSE’s Fannie and Freddie amongst others have mature systems which many conforming lenders use. The primary advantage of having an underwriting engine that is based on machine learning vs. the more traditional prescriptive rule engines is that machine learning-based systems can learn and react quicker to the actual performance of previously originated and funded loans.

Within the conforming/conventional mortgage space, even though Fannie and Freddie guidelines and systems influence about 80% of the market, machine learning-based underwriting can still play a role for lenders that have a custom underwriting overlay and serve as an additional decision point.

The impact of machine learning as a means of risk management and evaluation would be much greater within the 20% of the market that is non-conforming where GSE guidelines do not apply. This is relevant in the rise of the gig economy and rising volumes of non-conforming loans where there is an increase in sources of income from unconventional sources.

The benefits of machine learning to evaluate risk can be applied to other domains as well – consumer loans, student loans, auto loans, credit cards, auto insurance, home insurance, etc.

Default Expense Management

When borrowers have financial difficulties making payments, servicers typically operate on two parallel tracks – payment assistance to the borrower and collateral protection(the home). For example, sometimes borrowers abandon the property in which case the servicer is usually on the hook for adhering to the local ordinances – grass, trees, no broken windows, roof, fence, etc. Servicers incur upfront expenses around inspection and preservation and claim these expenses from the entity that owns the note (Fannie and Freddie for example). The guidelines around what is reimbursable vs. not are complex and are changing all the time. The risk for servicers is incurring expenses that will not be reimbursed. Machine learning can be used to serve as an additional decision point to review and flag potential expenses before the expense is incurred.

Reserve Forecasting

During natural disasters, mortgage servicers assist impacted homeowners in a variety of ways including payment forbearance and loan modifications. Machine learning-driven models can help mortgage servicers gauge portfolio impact and forecast capital needs to ensure that servicers have the necessary reserves to (figuratively) weather the storm.


Guiding Your Success

SpringML was founded in 2015 and is a Google Cloud Partner. We have more than 250 implementations. Don’t let the ML in our name throw you off! We have chops in multiple areas and are happy to partner with you on your cloud journey for a variety of use cases – cloud migration, data management, analytics, visualization, machine learning, contact center artificial intelligence, and document processing to name a few. Contact us, and let’s discuss how we can partner.

Thought Leadership