Machine learning: Turning data to information, information to insights

Science has brought us many inventions which have been used in various ways for the benefit of the humankind. One of them is Machine learning which is the buzzword for this decade. It has been doing rounds all around the business. When it comes to taking decisions, prediction plays a great role in taking apposite decisions which might have a huge impact on the business. Computers have helped us in a great way by improving our ability to take decisions. 

But what if we could improve the ability of computer to take decisions?  

In 1959, a new invention took place in the field of computer science which has now taken a large form through a series of continuous developments. We know it commonly as Machine learning. As the name suggests, Machine learning gives machine the ability to learn based on some algorithms. In short, Machine learning gives machine the ability to improve its own performance.

Machine learning is closely related to computational statistics which also focuses on prediction making through the use of computers.

Machine learning and Artificial Intelligence are closely related and perhaps this is the reason that they are often used interchangeably.

Artificial intelligence on one hand seeks to re-engineer the attributes possessed by human. AI refers to the ability of a computer to think like a human or mimic a human mind. It tries to match the logical skills of a human. Virtual video games, self driving cars and Siri ( the personal assistant of I-phone) all are applications of AI.

Machine learning

Machine learning is a technique which involves algorithms or models that learn patterns in data and then predicts similar patterns in new data. 

Machine learning as defined by Tom Mitchell, Professor at Carnegie Mellon University is as follows:

"A computer program is said to learn from experience 'E' with respect to some class of tasks 'T' and performance measure 'P' if its performance in tasks 'T', as measured by 'P', improves with experience 'E'."

In simpler words, if a computer program can perform a task based on some past experience then it is said to have learned something from past experience. This is quite different from a program which can perform a task because its programmers have already defined all the parameters required to perform that specific task.

For example: A computer program can play a game of tic-tac-toe if it has been programmed accordingly by a programmer with a winning strategy, however, a program with no predefined strategy and which has been programmed only within a set of rules will have to adapt by playing a lot of games until it understands a winning strategy.

This is true not just for a game but also for programs which perform classification and prediction. Classification is the process of assigning items in a data set to classes or categories. Prediction which is also known as regression is the process where a computer predicts the value of variable based on past values. 

People feel amazed at the pace machine learning has rose up in the recent years however, there is big reason behind the sudden rise of machine learning. Machine learning is not something new, however, the only reason which gives machine learning a boost is the presence of huge amount of data now a days which was not present earlier.

The major factors which have contributed to the resurgence of machine learning are:

1. Data mining
2. Bayesian analysis
3. Inexpensive storage

Machine learning can be divided into three categories:

1) Supervised learning

2) Unsupervised learning

3) Reinforcement learning

Supervised learning: Supervised learning is a method of learning where the data is tagged with labels and the machine predicts the outcome based on similar past events.

For example: prediction of the flag of a country when flags of the other country with country tags are mentioned.

Unsupervised learning: Unsupervised learning is a method where the machine is trained using a data set which does not consists of any labels.

For example: The auto prediction of google search engine is based on this way of learning.

Reinforcement learning: Reinforcement learning is based on behaviorial psychology. This sort of learning can be used in economics and game theory.

There are many methods to implement machine learning which are widely being used by data scientists.

Future aspects and applications

Machine learning can be used in wide range of applications to enhance which encompass all the aspects of business decision making i.e Finance, Marketing, HR and Operations.

1. Machine learning will bring innovation in accounting thereby reducing the repetitive task done by the accounting professionals which will enable them to focus on more important aspects that affect a business.
2. Machines can be taught  to predict the future revenue of the company.
3. We already have seen the potential of data visualization in business. The next step in data visualization is perhaps Automated data visualization which the companies would be foraying into by trying to choose right widget for displaying machine learning results in visualization softwares.
4. Companies are often left thinking the reason an employee quit a job while other continued to work for the organization. These questions can be answered with the help of variables like tenure, wage, time in current role etc which can be put into algorithms which would then be able to find outcomes that would otherwise be very difficult.

 References:

1. http://www.cs.cmu.edu/~tom/ 
2. https://www.forbes.com/sites/bernardmarr/2016/12/06/what-is-the-difference-between-artificial-intelligence-and-machine-learning/#728e84f12742 
3. https://www.google.co.in/amp/s/www.forbes.com/sites/bernardmarr/2017/07/07/machine-learning-artificial-intelligence-and-the-future-of-accounting/amp/ 
4. http://www.hcamag.com/hr-news/the-future-of-machine-learning-and-human-resources-236576.aspx
5. http://bigdata-madesimple.com/wp-content/uploads/2016/02/un-supervised-learning.png

Note: The views expressed here are those of the author's and do not necessarily represent or reflect the views of DOT as a whole.