Getting started with Machine Learning - Part 1

From previous article, we had a glimpse of what Machine Learning is and how some special software programs called "Machine Learning Algorithms" learns from data. Every machine learning algorithm undergoes rigorous Training and Testing process till its good enough to predict an answer that is as accurate as possible. 

Now, with this basic idea of Machine Learning and Algorithms, let us see how can we prepare and get started.

To begin any journey we need some basic things required for that journey, a plan (of how we travel), and most importantly the destination. Likewise, for Machine Learning, what is the motive for us to do this, what kind of problem we are trying to solve, and do we have enough "DATA" to explore about that problem and find patterns and solutions from it. 

Picture: www.bigdataframework.org

Lots and Lots of Data is the backbone for Machine Learning. This data can be anything from numbers, texts, pictures, videos, audios etc. We just need appropriate information for the kind of problem we need to solve. 

Once we have the necessary data, we then have to decide which approach will provide best solution for the problem we try to solve. There are various approaches in Machine Learning that are best suited for different kinds of problem. Some of them are

• Classification
• Regression
• Clustering
• Dimensionality Reduction, etc.

Don't be afraid of these new jargons. we will get it one by one as we move along. 

Lets see an example for "Classification" problem. Traditionally when a patient go to doctor for a severe and chronic pain, the doctor looks into patient's Xray images to diagnose the reasons for his/her illness. This diagnosis procedure may take some time and treatment starts late. But imagine, what if we had a ML algorithm that can quickly and precisely identify a tumor in a Xray image? 

picture : Shutterstock 

We provide the algorithm with thousands or millions of Xray images with tumor and without tumor. From these images the algorithm identifies the patterns and tunes itself to differentiate the images with tumor from the ones without it. Here the algorithm "Classifies" the affected and non affected images, which considerably helps to lower the diagnosis period and helps patients to get treatment as soon as possible. 

There are many examples for Classification problem. 

• classifying SPAM mails from non SPAM
• classifying different kinds of animals and birds
• classifying pictures of different persons
• classifying fraudulent money transactions from normal ones etc.

for all the above examples, it will be difficult for humans to go through millions of images or entries and segregate them accordingly. But once the Classification algorithm is trained enough, they can handle this monotonous task pretty well.

Being said that, in next post I will explain other approaches like regression, clustering, etc. with example and then we move to next level and see what are the different types of machine learning branches.

Machine Learning is such a vast topic and grasping the concepts can be overwhelming. For that reason lets us proceed one step at a time and try understand every detail as we proceed.