How to start Machine Learning journey: A step-by-step Guidance

In this growing world, technical achievements are playing a vital role. We must have heard that technology has started to solve some of the old unsolved problems quite easily. Machine Learning is one of such technologies which is solving many problems across wider verticals. ML is helping doctors in the medical industry, armies in the defense, farmers in the agriculture domain, scientists in their innovation, and many more.

Machine Learning and Artificial Intelligence are some of the hottest topics that can ensure a brighter future for any individual in the upcoming decade. As per Indeed’s post, Machine Learning Engineer is the top-rated job among the best jobs of 2019. The average salary of an ML engineer is $146,085.

But the technologies like AI and ML are new in industries; hence becoming a machine learning engineer or artificial intelligence engineer is not so trivial. One can easily find the content over the internet or take courses on some online platform to start the ML journey.

In this article, we will be sharing those steps which actually were followed by several industry professionals working in the ML industry. So we can say that It is actually verified. These steps may differ from conventional methods, but we can assure you that these steps work.

Step 0: Learn the basics of Python

Python is the most commonly used language in the field of machine learning and deep learning. Hence we would recommend getting familiar with the basics of this programming language. To learn Python, you can pursue either of the two options :

Programming for Everybody (Getting Started with Python).
Google Python Class

Note: There are other programming languages like R, Scala, Julia, but I am emphasizing Python because of six major reasons shown in the image below.

6 reasons why we should go with python for machine learning

Step 1: Try to find answers to some basic questions related to Machine Learning

To start the learning of any new technology, we must know answers to some basic questions.

What is that technology about?
When that technology started?
Why that technology even exists?
Where can we use that technology?
How does that technology work?
What are the scopes of that technology in the coming decade?

You can find the answers to these questions inside two of the blogs here in which the first four questions have been answered, and here the fifth question is answered. The more you dig, the more enriched your learning would be. This step is essential if someone is interested in making his/her career in this domain.

Step 2: Get yourself familiar with the Basics of Mathematics required for Machine Learning

Note : You are absolutely free to keep this step in parallel to later steps, as starting from maths may be tough/boring psychologically and we personally believe in the philosophy of “Learning by Doing” .

To become an expert in Machine Learning, one should know the basics of some most frequently used maths inside Machine Learning. Concepts like Linear Algebra, Probability, Statistics, and Calculus can be considered the machine learning basic building blocks. You need not master these concepts as only some parts of these concepts will be useful in problem-solving. You should learn these concepts not only from the theoretical angle but also from the python libraries angle. This means you must get familiar with the python libraries that actually help us in such mathematical calculations.

Below are the links for the free courses which will introduce you with,

Python libraries that are very useful and must to explore are :

Numpy
Scipy

With these libraries' use, one or two code lines can implement hundreds of mathematical implementation lines. Not only this, these libraries optimize the calculations a lot so that our algorithms perform more than 10X times faster.

Step 3

We have divided this step into two parts where both the parts work in parallel,

Part 1: Get yourself familiar with the concept of Data, Analysis, Manipulation, Filtering, and Visualisation

This step is one of the most crucial steps in the field of Machine Learning.
There is one famous saying that,

Ability of extracting the useful information from the given dataset is the key to differentiate between an average and better Machine Learning engineer.

The machine learning model works best with pre-processed and meaningful data. In this step, you should also explore some python libraries specially designed to help you in the pre-processing steps.

Python Libraries that are helpful for this step would be :

Pandas [Option 1 (preferred) , Option 2]
Matplotlib ( Used to visualize the dataset )
Seaborn
Scipy

One can go through some online courses to know more about data-preprocessing things. Courses that you can have a look at to master the data pre-processing steps are :

But keep in mind that the depth of this field is also infinite. But you are here for machine learning, not for data analysis or data science. Hence, we will explore things as per our requirements only.

Part 2: Choose any open source dataset and apply the learnings of part 1 in parallel

As we said, there is no point in learning things if you did not practice that learning. So we would recommend finding some open-source datasets that you can find here. Download any dataset and implement the learnings of part 1.

Task: For example, download the COVID-19 dataset, clean it, find the region with a maximum number of Covid cases.

Step 4

This step also includes two parts, where part 1 is again focused on learning, and part 2 is focused on testing the part 1 learning.

Part 1: Introduction to Machine Learning

At this step, you are fully ready to start your journey towards machine learning. You can start going through any basic course.

Machine Learning by Andrew Ng ( This course is good for understanding but uses MATLAB/Octave for implementation. )
Machine Learning Crash Course by Google
Machine learning with scikit-learn

There are many other resources that you can find on the internet. Still, as we focus on the language python because of the tremendous support available on the internet, I would recommend learning ML with scikit-learn. Always try to be in the depth of the concept and try to find the answer of What, Why, and How of any concept or algorithm.

Python library that would be great to explore

Scikit-Learn

Part 2: Decide a Problem Statement that you want to solve using Machine Learning

This step is a must if you want to be an expert in the area of Machine Learning. We have seen many professionals who have good certificates, but they cannot solve the problem.
Hence we would strongly recommend solving at least one problem statement available in the free domain. This would give you enough confidence to solve the related problem in that area. You don’t need first to complete the full course and then try the problem. It’s the older way which is slow. Pick any one problem of simple complexity and solve it step by step with the internet's help. Keep the course going on in parallel.

After following the above steps, you would become a good ML engineer with hands-on. Try some moderate complexity problems and then the harder ones. If you really want to become passionate, collect some research papers in the field of Machine Learning and try to implement them on your own. It will make your experience better and make you confident enough. Try solving one problem based on each algorithm covered in the course.

complete process of machine learning works

The overall process for the machine learning model development can be seen in the above image. Model deployment can be removed from the picture, but it's always good to build something completely and offer it as a complete solution.

Step 5: Learn Deep-learning and Neural Networks

Repeat the same strategy of Step 4 and follow the philosophy of Learning by Doing.

Deep-Learning courses that are really great

Deep-learning Specialization by Andrew Ng

We also have a completely different way of learning the Deep-learning concept, but that also includes the above course. We would share that pathway in a separate blog.