mechevarria / loan-default-prediction

Extended analysis with Eureka AI

Default Loan Prediction

This is an end-to-end project building a classification model to predict loan default using the Berka dataset.

This fork does additional analysis using the EurekaAI platform. Please contact SymphonyAI to get access to both the SDK and Eureka Workbench

Fork Changes

• Added scripts to install MySql in a docker container and load using the mysql client in the container

MySQL Server Setup

• Create data dir and run container with the local script

./run-mysql.sh

• You can shutdown the MySQL server with

docker stop mysql

Import the data into MySQL

• This script will copy the data files and scripts to the container and import using a local mysql client

./import-data.sh

Setup

• Install miniconda as an environment wrapper
• Create an environment

conda create --name sai python=3.7
conda activate sai

• Install the EurekaAI SDK

The SDK tar.gz installer is not included in this source

pip install ayasdi-sdk-3.0.0.7.tar.gz
pip install python-dotenv

• Setup additional libraries

conda install ipykernel pandas seaborn scikit-learn mysql-connector-python

• Create a file called .env in the root directory and put your EurekaAI platform credentials in it along with the Eureka API backend

Do not check a .env file into source control

EUREKA_USER="[e-mail]"
EUREKA_PASS="my_password_goes_here"
AYASDI_APISERVER="http://platform.ayasdi.com/workbench/"

Links to Resources

• My full post on Medium about this project

• Description of the Berka dataset

• SQL query to import the data into MySQL database

Notebook

• Exploratory Data Analysis (EDA): html and jupyter notebook.
• Modeling: html and jupyter notebook.