Hands-on with Machine Learning | A practical account

September 9, 2019

Hands-on with Machine Learning | A practical account

As the previous Financial Manager of Swipe iX, I had the fortune of being part of their culture of encouraging innovation and willingness for tackling the tough problems. I got to learn from some of the most amazing developers, project managers, designers in the business. 

Through various conversations with staff and management about opportunities to gain exposure to the vast world of Tech, I stumbled upon an initiative called Explore Data Science Academy. With Swipe IX’s blessing, I pursued a year of intensive training in the art of Data Science with which I’m currently busy. What follows is a Professional Accountant (S.A), ex-Senior Auditor’s account of hands-on experience with practical Machine Learning solutions.  

Why is Machine learning important?

This has to be the biggest case of fear of missing out that has ever taken the corporate world by storm. Every Director from CEO to CFO is feeling the pressure from various board members and stakeholders. Everyone knows it's significant, rightly so, but why is it important? How does this take the business forward? How does it add to your bottom line? What is it good for?

To put it simply we learn constantly - we're limited however in that we can only take in so much in a finite amount of time. We're most days all in desperate need of more sleep and more time.

What if we could do away with sleep and kick time in the teeth? 

I've tasted some of Machine Learning's absolute power. 

During my sabbatical, I've received immense exposure to various techniques. The importance of Machine Learning lies in the benefits:

  • Advanced insights into big data all brought about by a computer program.
  • Predictive capabilities that stun you as to their accuracy.
  • Automation that will make you rethink the need for manual labour.
  • Speed.

How does this take the business forward?

It's a trick question - it should be: How does this NOT take the business forward?

The biggest challenge is not technical knowledge, although strenuous, nor is it hardware requirements. It's finding the correct partner. 

Simply it takes the business forward by:

  • Automating manual tasks, reducing the capacity needed.
  • Crunches the big numbers too large to imagine and provides critical information for business decision-makers.
  • It predicts the future with fear-inducing accuracy.
  • Brings about decision making based on statistics, removing human error.
  • It does so fast.

Training a machine learning model via Python to be able to predict whether someone is an introvert or extravert.

Creating a visual high-level overview of data-set composition to determine the balance in data.

How does it add to your bottom line?

We recently completed a project making use of Machine Learning - Classification.

We speculated that we could help save the insurers  Millions with a Machine Learning Model that can predict fraud simply by analyzing claim statements made by clients who are potentially trying to defraud the industry. 

This model would pick up patterns from the data and provide a suspicion indicator. 

Why would this revolutionize the industry? Currently, the South African Insurance industry sufferers reported fraud losses of R 1 Billion annually. It is estimated to be up to 3 times more than reported.  There are systems in place that predict and preventing fraud losses. However, an inherent aspect of Information Technology is: Computers think Numbers.

So how can a computer judge fraud based on the written text? You've guessed it - Machine Learning.

Our program will provide a suspicion indicator that can patch into the existing, actuarial models, and as it's proven to be effective reliance on it can be increased. This would then give the actuarial models not only the capabilities to catch fraud out on a quantitative level but also on a qualitative level.

If we made a 1% Difference in the industry we will save the insurers R 30 Million a year. To develop such a program would only amount to a third of the savings. It will then continue to service the industry indefinitely and being a machine would not ask for a raise at financial year-end.

What is it good for?

Einstein famously said: "the true sign of intelligence is not knowledge but imagination".

Currently, Machine learning is being used widely for:

  • Image Recognition
  • Speech Recognition
  • Medical diagnosis
  • Statistical Arbitrage
  • Learning associations
  • Classification
  • Prediction

We've invented these tremendously powerful building blocks. Uses limited only by our imaginations.

Take the following with you - ensure that when you want to step into this territory with your organization that you find a partner that is technically savvy and imaginative extraordinary. 

Jonathan Aspeling

Guest Author

Swipe iX Newsletter

Subscribe to our email newsletter for useful tips and valuable resources, sent out monthly.