Bankruptcy Prediction

Team Name

Rookie Hackers

Team Members

Project Description

This project is a part of GovHack 2019 Challenge. Challenge details and datasets have been provided by Australian Financial Security Authority (AFSA). Challenge aims to make it easier for people to tell AFSA their income and assets when they become bankrupt.

Data Story

Background

The Australian Financial Security Authority (AFSA) is responsible for regulating the personal insolvency and personal property securities regimes.

What is bankruptcy? (Source - AFSA Website)

Bankruptcy is a legal process where you're declared unable to pay your debts. It can release you from most debts, provide relief and allow you to make a fresh start.

You can enter into voluntary bankruptcy. We refer to this as a debtor's petition. It's also possible that someone you owe money to (a creditor) can make you bankrupt through a court process. We refer to this as a creditor's petition. Bankruptcy normally lasts for 3 years and 1 day.

Bankruptcy trustee
When you become bankrupt we appoint a trustee. A trustee is a person or body who manages your bankruptcy.
This can either be the Official Trustee (AFSA) or a registered trustee. You can also nominate a registered trustee of your choice.

Debtors Obligations

When you are bankrupt:
you must provide details of your debts, income and assets to your trustee.
your trustee notifies your creditors that you’re bankrupt - this prevents most creditors from contacting you about your debt.
your trustee can sell certain assets to help pay your debts.
you may need to make compulsory payments if your income exceeds a set amount.

About the data

Personal insolvencies include bankruptcies, debt agreements and personal insolvency agreements. We have provided data on personal insolvencies for calendar years 2007 to 2018. Data for 2007 are from 1 July 2007, and data for 2018 are to 30 June 2018.

When you enter into a personal insolvency, you must complete a statement of affairs. That form asks for a range of information, including your income, debts and assets. We have provided debtors’ responses from this form. We have published this information as the debtor initially reported it, irrespective of subsequent checks and revisions.

More information about the dataset can be obtained by visiting the link below:

Bankruptcy Challenge Dataset

Problem Statement from AFSA

Validate Income and Assets declared by Debtors:
When people go through personal insolvency, it could be very stressful time for them. When they complete the statement of affairs they might forget to include certain details e.g. other sources of income, all assets, etc. AFSA wants to understand how we can predict the debtors income and assets using social, employment and other insolvency related information provided by debtors.

Using Debtors information predict Income and Assets over 3 year period:
When people file for bankruptcy they are usually bankrupt for 3 years so AFSA wants to understand when their circumstances might change. Or within limits, AFSA might be able to sell their assets to repay their creditors.

Example
John is 44, a builder and is currently not working because of health problems. If he was bankrupt, how would AFSA tell if he gave AFSA accurate data when he declared bankruptcy and whether his circumstances will change over the next 3 years e.g. getting another job, buying houses or shares etc?

The Hack
As a part of this hackathon, we will fundamentally focus on Problem statement 1. We will use AFSA dataset to build a Machine Learning model to predict debtors' income category based on other socio-economic attributes described in the dataset.

This document goes through following steps:
1. Importing Dataset and Relevant Python Libraries
2. Data Cleansing and Exploratory Data Analysis
3. Feature Engineering
4. Predictive Model Building
5. Validating Model Accuracy

Now let's do the deep dive...