Project Description
Our product is called InSOLVE.
We used machine learning and/or artificial intelligence, to predict non-compliance during personal insolvency/bankruptcy.
We trained a neural network to sort through the data that we used, and was able to create a website that can showcase basic insolvency data via database lookups, but also more technical temperature matrices using Mathematica, which has been programmed via MATLAB neural network matrices.
The Australian Financial Security Authority (AFSA) would find this information important in investigating people after they have filed for bankruptcy.
Regional communities like Rockhampton can improve their economic resilience through using Artificial Intelligence like Neural Networks to make sure that integrity is kept. While it is not illegal to go into bankruptcy itself, not complying with insolvency orders afterwards is an issue. Rockhampton and Yeppoon can improve a lot by making sure that at risk members of their community can get the extra help that they need, and the indices that we calculated can help do this ahead of time.
Try putting in your details and see if you are at risk of non-compliance in the case you ever become bankrupt?
Data Story
We found common triggers among non-compliant records, for example location, job, family status were all variables.
In the bankruptcy dataset, the criteria that we looked at was;
*Calendar Year of Insolvency
*SA3 Code of Debtor
*Sex of Debtor Code
*Single/Couple
*Dependants
*Debtor Occupation Code (ANZSCO)
*Cause for Bankruptcy
*Business Related Insolvency Code
*Debtor Income Level
*Primary Income Source Data
*Unsecured Debts Levels Value of Assets Levels
Using data from Australian Financial Security Authority (AFSA), we sought out to see if there were common triggers in these cases. For example, those who are living in Rockhampton who are single & without dependants are a large group of insolvencies, while those who are living in Rockhampton while being a couple and having dependants are more likely to not comply with their insolvency notices compared to others in Rockhampton.
We have compared census data from the 2016 ABS dataset which includes Rockhampton's populations with respect to age and sex.
We utilised MATLAB packages trained a neural network using our datasets, then the resultant code was implemented in Mathematica.