What is Dimensionality Reduction?

When we have all the numbers, we will need to get rid of the least important ones. This is a step know as Dimensionality Reduction. To do that, we will need you to choose how do you want the reduction to be made. We will present you with 3 options: Principal Component Analysis, Linear Discriminant Analysis and Correlation. We will provide an explanation on how each of those work so you are able to make an informed choice. These 3 algorithms will all do the same but in different ways: take all the variables that we got from Radiomics, and discard the ones the influence of which is negligible. This way, instead of having to train a machine that takes into consideration a thousand different little things, we will train a machine that will look into the 10 or 20 truly important variables. This step is important because, when the machine is getting trained, it will take a much shorter amount of time if we analyse a smaller amount of data. If we were to train it with every single variable and feature that Radiomics takes from the magnetic resonance imagings, it would take way too long. As we mentioned before, the key is to find a balance between the amount of time it takes to train the machine and the amount of variables we let go. It's a sort of compromise.