Customer Churn Shiny Application
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
data
images readme + rename rmd Jan 10, 2018
model
rsconnect/documents/customer_churn.Rmd/shinyapps.io/jjallaire
.gitignore cleanup repo Jan 10, 2018
README.md
customer_churn.R
customer_churn.Rmd

README.md

keras-customer-churn

This repository contains a Shiny application that demonstrates the use of the customer churn model described in the Using Deep Learning With Keras To Predict Customer Churn blog post.

Using the Application

The application is deployed online at https://jjallaire.shinyapps.io/keras-customer-churn/.

To run the application locally, clone the repository then:

rmarkdown::run("customer_churn.Rmd")

The shiny application has three tabs:

  1. Customer Scorecard---Analyzes a single customer at a time. The keras model is used to return the probability of customer churn. The app then recommends three strategies to mitigate churn risk:

    • Main Strategy - Incorporates tenure, contract type, key services, monthly charges to recommend offerings that reduce churn risk
    • Commercial Strategy - Incorporates specific services that the customer may be interested in
    • Financial Strategy - Incoporates payment method recommendations to reduce churn
  2. Churn Facets---Analyzes aggregate churn by various features including type of contract, revenue, tenure and internet service. Drop-box filters are available to subset the data and drill into important customer segments.

  3. Correlation Analysis---The correlation analysis shows the features that correlate to churn, which is important for a global perspective of understanding what affects churn.

Training the Model

Use the customer_churn.R script to train the model used by the application from scratch:

source("customer_churn.R")