What is Machine Learning ?

Machine Learning is a technique to make computers understand patterns without being explicitly programmed. This is an area of Artificial Intelligence, a bunch of mathematical and statistical formulas trying to understand patterns hidden in data.

The area of study started with researchers trying to figure out how humans understand the world around them. It includes,

  • Data Collection: Humans perceive the data using five senses (eyes, ears, skin, nose and tongue) in the format that the mind can understand.
  • Human mind processes this data, understands patterns and comes to a conclusion based out of its past experience (i.e., objects it learnt, history of events happened or information gathered out of others experiences etc.).
  • When a new scenario arises, the mind will use the learnings/understanding to make necessary decisions.

Machine Learning tries to mimic similar process to make computer learn/understand patterns hidden in data. Machine Learning algorithms can understand and process only  the numbers which is a limitation, where as human mind can understand many things beyond numbers. The stepwise process of pattern recognition in Machine Learning is as below.

  • Collect digitized historical data.
  • Learn/Understand data patterns at high level using statistical data analysis and required transformations applied.
  • Convert this data into numbers (such as text, images, videos, etc..)
  • Feed this data to machine learning algorithm which will learn/understand patterns that is translated to a Machine Learning Model.
  • When a new record comes in, the model can make predictions (similar to human approach)

 
How Machine Learning (or Artificial Intelligence) can add value to business ?

Traditional way of Implementing Business Rules: In a traditional software company (where no Artificial Intelligence is used), a Business Analyst (BA) or Subject Matter Expert (SME) will understand business rules (or patterns), he/she will explain them to the programmer. The programmer will write a bunch of if-else conditions and loops to implement the functionality. When business rules change the programmer will have to depend on the BA/SME to perform the changes and the cycle of code, build, test, and deploy will continue until all the needs are met.

Here, the business rules understood by BA/SME are the basis for the whole system in making decisions. If for some reason, BA/SME is unaware (geographic location or changing patterns) of some scenarios, the system will fail to deliver best-in-class service in turn impacting the business.

Machine Learning way of implementing Business Rules: In a Data Science project, we consider all the above traditional approaches, on top of it Machine Learning or Artificial Intelligence will bring insights that may have been ignored by BA/SME.

The stepwise approach

  • Perform statistical data analysis.
  • Statistics is all about making conjectures, hence we will get some clues on data problems. Take help form BA/SME to confirm and correct data problems.
  • Now feed this data to Machine Learning/AI algorithm and train.
  • The model is ready to make predictions or suggest recommendations.

Here, we consider BA/SME understanding, on top of it we are using data to bring out patterns (instead of programmer coding if-else conditions). The advantage of using Machine Learning approach is that it will bring out all hidden patterns in data. Especially in this data rich era (lots of digitized data), using AI/ML will bring lots of benefits to businesses.

The building blocks for succeeding in AI is to have a strong hold of basics in ML. There is a necessity for every individual associated to Data Analytics (especially inclined towards Predictive/Cognitive Analytics) gain this knowledge to come up with Innovative ideas and connect at the grass root level.