How to choose variables to perform Exploratory Data Analysis

Question

I have a dataset with about 110 variables. I have a target variable and I want to do an exploratory data analysis to find out what factors affect this target variable.

In such scenarios when we have a lot of variables,how do we choose which variables to analyze? Should we consider all the variables wrt to target variable or do we choose variables based on domain knowledge?

I am a newbie to DataScience,Need some guidance on how to proceed with analysis when we have large no of variables in our data.

Answer 1

There are a couple options:

1. Iterate over all input features & calculate it's correlation to the target variable. Gather all these numbers and sort them by absolute value. Take the top 10 or 20 as a chunk to start investigating these features with more attention (absolute value because you care about strong negative correlations just as much as strong positive correlations)

2. Train a simple Decision Tree on the inputs mapping to the output. Once the decision tree is trained take a look at the feature importance(s) that the decision tree uncovered - begin your investigation here. You can repeat this process with a linear regression too.

3. Plot all 1-to-1 plots of all input variables to the target and manually look through them (this takes more time as you need to look through as many plots as you have input variables - but it will give you a good understanding of your data once you go through it all)