This group looked at attrition data from IBM's HR department to try and find features that correlated highly with attrition.
What was good:
Just overall the layout and organization of the report was very nicely done. The logical outline of data exploring, data modeling, and then the three models each with description, results, and analysis, is very clear and direct which made it easy to read.
The methods used in the analysis went beyond the scope of the class: SMOTE for addressing the unbalanced classes, SVC algorithm for SVMs, telescopic search for model parameters. These were very thoughtful choices to make in analysis and showed that substantial outside work and research was put into the assignment.
3.The data you chose to visualize was done well, visuals were very informative, the exploration wasn't excessive and you had a pretty cool correlation matrix to kick everything off.
What could be improved:
the table in results 4.2, could be formatted a little better, there's a lot of white space there due to text in cells in the first column crossing over to the next lines.
Maybe in the introduction you could have more about the context of the problem you're solving. You start off with an abstract that describes your conclusions, but some more business motivation would be good (though this does come in at the end!)
I actually interned at IBM consulting over the summer and someone on my team was actually building a logistic regression model for attrition! She was framing the problem as a survival analysis problem, so maybe you could look into some of those techniques to improve your work.
This group looked at attrition data from IBM's HR department to try and find features that correlated highly with attrition.
What was good:
What could be improved:
Very nicely done!!