The machines, as we know of until recently, takes instructions from the human masters, and execute them. Humans have “codified” their learning into the machines, and these rule-based systems are changing now, with Machine Learning.
The goal of machine learning is to enable computers to “learn” through the “experience” of performing a “task”. But, why should the machines learn? The simple answer is “to help humans in various activities”. There are enough futurists, who predict a gloomy future for the homo sapiens, as “super machines” emerge. Keeping those fears aside, let’s look at how humans can benefit, through a simple example.
The Free Gmail Account
Do you remember the initial days, when you opened a free Gmail account? The allure of 1GB free storage and a sleek email interface, attracted not just you, but the spammers as well. Those days, you had to open the email, understand it was a Nigerian fraudster with $1m to spare, and then report it as spam. How often do you do it today? Rare, isn’t it? What has changed?
Machine Learning has taken over the job of detecting spams. Over the years, based on the feedback from the users – every time an email was marked as spam – machines learnt the “algorithm” to detect spams. Using those algorithms, they now identify spam emails, way before it reaches your inbox. Additionally, they keep improving the algorithms, to detect new spams.
Why couldn’t the humans do it? Imagine the amount of information you need to hold in your memory, and the processing power that is required to scan through all the emails that passes through Gmail servers? That’s serious heavy lifting! Better left with the machines!
So, what is Machine Learning?
From the Gmail example, you would have understood a few key things:
- Machines (computers) are employed to learn through the experience of performing tasks
- As they learn, they build their own algorithms
- These algorithms, they keep fine tuning, to improve the accuracy of future predictions
In its’ most simplistic form, Machine Learning methods are often simple extensions of well-known statistical methods. But, why statistics?
Let’s go back to the Gmail example. As an email passes through the servers, the Machine Learning program (or application) is performing these tasks:
- Scans through the email, looking for previously “known patterns” of a spam mail. These know patterns are based on the learning from the historical data
- Once it sees the patterns, it checks for what % of the new email, contains the patterns matching other older spams (a simple statistical formula to calculate probability)
- Say, if the new email matches 30% (cut-off) with the known patterns, then the email is sent to spam folders
Of course, the above example is a simplified version of what might be happening at Google. The larger point is, statistics is an integral part of Machine Learning field, and it is leveraged using modern technologies.
The formula for Machine Learning
If I were to sum up Machine Learning, my formula is:
Why use Machine Learning for Business Analytics?
A beautiful explanation on Analytics, comes from Customer Analytics:
“A good analytics solution tells a story of the past, present and the future, and is about finding hidden patterns in data to provide critical business insights and drive business change.”
Machine Learning programs:
- Provide the ability to crunch massive volumes of data, and learn from the past
- Enable ease of codifying statistical techniques (practically code is freely available for all commonly used statistical models in the internet)
- Use the past to learn & predict the future
Hence, people find Machine Learning a valuable tool for Business Analytics programs.
The Big Picture!
How is Machine Learning different from Artificial Intelligence? What is Deep Learning? There are enough confounding stories around Artificial Intelligence, Machine Learning, Robotics etc. As always, before closing, let’s look at the Big Picture.
Nvidia, a pioneer in this space, provides a clear definition on AI, ML & DL. Reading this blog is highly recommended.
Source : www.nvidia.com
In the upcoming blogs, I will cover how to get started with a Business Intelligence program.