add pipeline step for exploratory data analysis

[john-waczak]

Include

• Correlation matrix
• Mutual Information
• @AdamAker BARS stuff

Generate a report for a Dataset Overview

[john-waczak]

Add identifiers for missing, nan, inf, other types so that we can drop them later

[john-waczak]

Take as an argument a Table.jl or DTable.jl compatible table for this step. Also consider using OnlineStats.jl for processing large datasets over predefined chunks.

[john-waczak]

It will be desirable to specify lists of features and lists of targets to perform the exploratory analysis on. Then after reviewing the results, we can establish heuristics for which targets we want to train models for with which subset of relevant features.

[john-waczak]

For the feature importance evaluation, we should try to pick a model that will also allow us to estimate model sensitivity to features, e.g. MLPs, regression, etc...

[john-waczak]

sensitivity analysis in ML context