gvanzin / churn201 Goto Github PK

The Predictive Analytics Times has been kind enough to publish a series articles by Talent Analytics' Chief Scientist, Pasha Roberts. The articles focus is entirely on business and quantitative issues that impact Employee Attrition.

An online blog can only go so far, technically. So to foster conversation and learning, including learning by us, we are putting more content here on GitHub. We believe that GitHub is an important and useful platform for collaboration between data researchers, enabling direct sharing of code, data, ideas, and results.

In this series we are posting the R code used to generate graphs, models, and insights. In a way, this model is the unified engine for this Churn 20x Series. We have put this code, currently over 500 lines of R, online so that other researchers (and clients) can download, experiment, and engage. If you don't like our Weibull distributions, you can swap in a Log-Logit or whatever you want. If you want to create a U-shaped cost curve, go ahead.

• Build plots
• Modify employee cost/benefit models.
• Based on PA Times Article: Churn 201: Calculating Employee Value (2/20/2013)

• Run interactive simulation of hiring costs
• Calculate Expected Value of Hiring
• Preview of sensitivity study
• Based on PA Times Article: Churn 202: Good and Bad Churn (4/1/2013)

(coming up)

(coming up)

What is the employee analogy to "Sleeping Dogs" and "Persuadables" in marketing churn?

(coming up)

(coming up)

Are false negatives or false positives more expensive?

(coming up)

Git is a non-proprietary source control system, entirely separate from GitHub. GitHub is more like a de facto central breeding ground for thousands of open software projects. The best way to use the system is to join and engage. It's free, but there is some SSL configuration to use it properly. Welcome to the new world!

If you are on GitHub, you just use git to "clone" a copy of the Churn 201 project to your local system. Barring that, you can simply click on the "Zip" icon to the right. You can even download the individual churn201.R and churn202.R files directly from the file manager above.

Once you have the files, let's say you have created an innovative new cost curve or dynamic system that you'd like to share. Or, maybe you found a bug and want to send your fix. You can share your these changes back to us with a "pull request". Or, you could "fork" your own variant of the project. If you find a bug, create an "issue."

Keep us posted and engage!

R is an open-source, free statistical platform that has taken the data science community by storm. There are free libraries for almost anything one would ever want to do, and it is increasingly seen as the reference platform for most work in this domain. We tend to treat R more like a statistically aware, vector-friendly programming language, than a pretty GUI system or graphing package.

The programs in this GitHub can be run either directly from the console, or in the free and powerful RStudio, or likely on other platforms such as RCommander. My personal preference is to use the console, but many users will want something more visual. The interactive, slider-based model in Churn 202 does use the "manipulate" libraries that are only found in RStudio.

Running the code is easy:

• Open R, R Studio, or whatever R platform you use.
• Set the working directory to your downloaded/cloned directory with setwd("/run/dir/name")
• Load the libraries and program with source
  • source() the relevant program, such as source("churn201.R") or source("churn202.R").
  • This commands loads the relevant libraries and programs into memory.
  • Success is silent, but in RStudio you will see objects load into the top right panel.
• If you are missing libraries, you will need to install some packages.
  • For example, Churn202 uses the following packages: ggplot2, gridExtra, scales, and manipulate.
  • In RStudio, use the Tools>Install Libraries menu, or find the same option on the lower right panel.
  • From the console, simply run install.packages().
• Follow the instructions on the companion page, Churn 201 or Churn 202:
  • Run short commands as noted
  • Explore, view, modify objects
  • Create and modify plots
  • Modify the code, re-source, and re-run

As new blog articles come out, we will publish additional code to support the new ideas.

Copyright © 2014, Talent Analytics, Corp. All Rights Reserved.