How can I choose the best machine learning algorithms from all kinds of algorithms?

Question

When I want to find a model for my data set, I find that there are lots of algorithms that I can use. I know how to minimize selection choices by separating supervised and unsupervised algorithms and the purpose of the problem I am trying to solve.

But after that, there are also lots of algorithms to choose from, even in the scikit-learn library that I currently use, and there are lots of algorithms that I don’t know. They might work better in my problem and also there are deep learning algorithms that are stronger than machine learning algorithms. After looking for them, I’ve got tired and a simple project cost me a whole two weeks, but I wasn’t satisfied with the result at the end either.

So, what should I do?

• Do I have to memorize all the algorithms in machine learning libraries, like scikit-learn?
• Or should I abandon learning machine learning algorithms and start learning deep learning?

Answer 1

There is a theoretical result called the "no free lunch theorem" which proves that there is no "best ML algorithm" in general.

• It's important to understand how an algorithm works in order to have a good intuition about whether it's suitable for a case. Without this one can only attempt different methods randomly by trial and error, it takes more time and more effort.
• From what you describe it looks like your learning focused on how to use available tools (i.e. run the algorithms). If you want to become really good at data science you also need a good theoretical background.
• Data Science is very broad, nobody knows everything because it's impossible. My advice is to focus on understanding one thing very well before moving on to the next topic. It's usually better to be an expert in a specific area than to have a shallow knowledge of a bit of everything.

there are deep learning algorithms that are stronger than machine learning

Technically deep learning methods are also machine learning.

Or should I abandon learning machine learning algorithms and start learning deep learning?

In my opinion it's better to have a really good understanding of traditional ML methods before moving to DL.