ML is a complex subject to get your head around, especially the terminology and mathematical methods employed in analytics. You will need to understand what your requirements are. For example, if its for fraud prevention then the classification technique can be used, if its for real time decisions Reinforcement Learning is generally used, if its for targeted marketing Unsupervised Learning and the clustering technique can be employed and so on.
This also means that different algorithms are more suited to different techniques, its a trial and error process. Its a complex subject but bit by bit it can be broken down into more understandable form that can help in getting a good overview of what its all about, you can then move on to the more complex aspects should you have the desire to. have a look at the cheat sheet below from SAS it shows some of the dependencies within a flow trying different techniques.