machine learning algorithm summary

Which are the most important machine learning algorithms? Anyone who has been part of this domain must have faced or posed this question at some point of time. I too am asked this often.

First things first – there are no winning algorithms. For different circumstances, different algorithms, even though they might be designed for similar outcome, result in differently oriented output. Depending upon what you want with your data analytics, an algorithm might be better suited to you than the others for that situation. Size of the data set also plays a key role in determining the model to apply. Also, iterations in existing algorithms, thereby increasing their relevance to myriad of applications, is also common. What is important to remember is that simpler algorithms aren’t bad or obsolete. So, my suggestion is instead of searching for the best algorithms, one should focus on gaining awareness about fundamentals of different algorithms and their applications.

Most of us familiar with the subject would recall that in 2006, IEEE Conference on Data Mining identified the top 10 machine learning algorithms. That list is widely available over the internet, so we’ll not reproduce it here. What we’ll do instead is mention over a dozen algorithms, segregated by their application intent, that should be in the repertoire of every data scientist. For usefulness purposes, I am reproducing the list in an infographic format – that can easily go onto a wall!

For suggestions and improvements – please add your comments below.