This repository contains the framework for Deciphering Explicit and Implicit Features for Reliable, Interpretable, and Actionable User Churn Prediction in Online Video Games.

It consists of two components:

• Interface: Contains frontend code.
• Backend: Contains backend code. The counterfactual generation in the backend is inspired by and builds upon the DECE framework: https://github.com/ChengFR/DECE-research.

Interface Setup

This framework is based on Node.js and npm.

• To install Node.js, please refer to https://nodejs.org/en.
• To install npm, please refer to https://www.npmjs.com/.
• We recommend using nvm (Node Version Manager) for easy Node.js version management: https://github.com/nvm-sh/nvm.

Required versions:

node version: v16.15.0
npm version: v9.6.0

Installation

Once the environment is set up, install the required node modules by running:

npm install

Backend IP Configuration

In interface/config/index.js, replace YOUR_BACKEND_IP with the actual backend IP:

dev: {
    env: require("./dev.env"),
    port: 8050,
    assetsSubDirectory: "static",
    assetsPublicPath: "/",
    proxyTable: {
      '/cf': {
        target: `http://YOUR_BACKEND_IP:${proxyPort}`,
        changeOrigin: true,
        pathRewrite: {
          '^/cf': '/cf'
        }
      },
    },
    cssSourceMap: false
  }

Launch the Interface

To launch the Don’tGo interface, run the following command:

npm run start

Backend Setup

To run the backend, execute:

python app.py