I am starting to learn Exploratory Data Analysis for ML and I have some doubts that I hope to find some help here. First of all, I have been testing some classifiers such as Logistic Regression and SVM for breast cancer classification. The dataset in which I have been working on is:
In that site there are two datasets in which I have been working on and I have some questions about EDA that might seem simple. By the way, my background is pretty vague in statistics.
The questions that I have is about the importance or the relationship that has the use of boxplots for EDA, this before applying a ML model. For example, I got a couple of boxplots like this:
So it would be safe to say that in the case of the samples that belong to malignant cancer data, that the mean is greater than for the normal or healthy cases?
I also observe that on the bare_nuclei the dots, could be probably outliers; is that so? The data has been normalized and the missing values were replaced with the mean of the values. In this case how do these outliers could affect my classification? Should I perform a R2 with the output value (benign or malignant sample) to see if they are correlated or not? So in case they do not present a strong correlation can I dismiss this feature?
Hope that they could help me with these questions. I have found plenty of tutorials of ML on the web, but about EDA there is not so much information. I would like to ask for material about this topic, but I think is not allowed because of the policies of this site.