Challenge #127: US Grand Prix Lap 2 - Employee Retention

[Reesetrain2]

All,

Really interesting!

-I was able to get the challenge done in about 10 minutes, but I do use the predictive tool a great deal.

-After I was able to get everything to work, I went back and started digging around in the summary reports.

-The Branches for the Decision tree are at 441 and then 1187 days of employment, performance score: Exceeds expectation or probation vice all others.

-In short this means that employees either stay just over a year, or just over three years, as there is no consequential difference according to the decision tree.

-As far as performance goes, the employees that exceed expectations leave and the employees on probation leave, all others are essentially in the same bucket.

-Cheers!

Spoiler

 

Matt

[alacoume]

Not sure about this one...

[DaisukeTsuchiya]

Spoiler

 

[harrame]

I've got the correct workflow setup but I'm getting no employees with a Score_Terminated value above .5 for the Decision Tree model. I downloaded another person's solution and their workflow, too, did not work for me. Is this a version issue? Or a user issue?

Spoiler

[DataNath]

My solution to #127. Same issues as others but the working is there.

Spoiler

[Spurs]

Spoiler

[chandler-gjino]

attaching my solution

same as the others. 

decision tree was "more accurate" but not predicting any current employees to be at risk

[aiahwieder]

These predictive challenges are always heavy on the configuration . . . I still can't get to the right answer, but I'm at a loss to see where my workflow is failing.

[Charity_K_Wilson]

I've got the same issue.  Its like something in the coding changed for the Decision Tree model.  The Logit Model would produce some valid predictions, but the Decision Tree model no longer produces the original forecast.

[Charity_K_Wilson]

Well, it's completed.  But the solution was unattainable.  As a few have noted, there was some version update.  

I tried both the Decision and the Logit modeling.  The Logit forecast was different, and forecasted different people.  (Can we talk about model validity if they differed this much!?)

Like a few others, I also tried downloading and running previous solutions.  When I downloaded them I was unable to generate the anticipated answer.  Its a back-end glitch.  Everything is coded right.

Spoiler