INTRODUCTION TO MACHINE LEARNING (ML)

credit:www.crowdanalytix.com

Machine Learning (ML) has become a buzz word in 2016  and clearly means a lot in today’s technology. Looking at advancements from Google, Microsoft, Facebook, Amazon and other Companies to champion the use of this technology only tells you its importance.

The downside is that for new engineers it’s hard to comprehend and those who have tempted to develop it have found it cumbersome.

In simple terms, Machine Learning (ML) is a technology that allows the machine to learn how to do certain activities which couldn’t have been done with direct programing of the machine and will always change with future data. In most cases, these activities involve a lot of data, unidentified variables, non-linear repetitive (very dynamic repetitive activity) tasks and others.

Let’s get an Example

Imagine you had a website showing movies to users on the landing page, so for every user, a new recommendation is made. Programmatically, we can program a system to give them recommendations with a number of IF statements. But if the users taste or the trends change, do we have to alter the whole code base at every instance.

 This is where Machine learning (ML) comes in. The computer system can study and always learn our users and provide a not so accurate recommendation to them for the first time but with time and continuous learning, the accuracy can keep improving. Take note the accuracy on recommendations is not based on the recommendations themselves, thus even if tastes change due to the high accuracy we shall provide the more accurate recommendations and have more satisfied users.

How We Use It At Rainbow

At Rainbow, we use Machine learning (ML) in fraud detection. We aren’t going to disclose our system architecture here but we are going to throw some light. Being a company looking at transactional technology, we a bound to attract fraudulent behavior, but as humans we can’t assume and program different ways the system should capture them or respond in the eventuality. 

So we have built Machine learning (ML) models monitoring all system transactions that keep learning on how to better detect the fraudulent behavior, think of them as soldiers patrolling a building ready to stop anyone trying to trespass.

But come to think of it, fraudulent behavior should be avoided not detected, so we are also using Machine learning (ML) to detect early symptoms, before it occurs. Think of it , upon your first cough or temperature rise, we quarantine you and follow up closely since we have known the chance of you being sick is high, but then in case you finish the transaction well, we learn from that and next time, we shall be smarter.

Some of the readily available tools are TensorFlow from Google, SCIKIT & SCIPY .

That’s how we are doing it at Rainbow.More Machine learning (ML) Publications are on the way and probably with more code this time. Thank you and hope this is helping people better understand the technology of Machine learning (ML).