Okay. Principle component analysis. So, one side benefit is that it helps protect yourself from overfitting the model because all component analysis is used to reduce the dimensionality of your data. But it also can help you spot redundancies in your data, it helps you to find out if two or more variables are telling you the same thing. The output features are uncorrelated and this is a good thing. You can take a dataset with many dimensions and reduce it to the dimensions, to some number of smaller dimensions, okay. What's called the first principal component has the highest variance. So, it's the first set of features that have the highest variants in there and the second principle components. So, if you take, let's say, a dataset that has seven dimensions and then reduce it to two, which is easy to plot using the Python plot libraries, the horizontal axis, which is the principal component, has the highest variance and the vertical axis is the second principal component has the second highest variance. If you were to print in 3D, the third axis would have the third highest variance and so forth. So, there's a couple ways that principle component analysis can be thought of. One is as feature extraction, and the other one is as feature reduction, and it depends on your perspective. I look at it more like a feature reduction, but I can see how it might be used as feature extraction because you can take a 150 dimension dataset and reduce it down to say, seven dimensions, which are the features you really care about. That can be considered way to extract features. So, a word of caution. When reducing features is that you need to make sure that the performance of the model is in model, and it's results, and it's error, and enhance learning curves aren't adversely affected by reducing the dimensionality of your data. You need to carefully evaluate each variable measuring it's usefulness. It's not a magic bullet. It may help you, it may not help you, and only you can make that decision. If you go down the PCA route and it's not working for you, then you're going to have to look at other techniques that were mentioned previously and there's probably even more techniques than the ones I had listed.