| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 2.2 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
The current society is volatile, influenced by macro social, economic, geopolitical, and natural phenomena
that have a global and deeply interconnected impact. As a result, as unpredictability increases, access to information and decision-support tools becomes increasingly vital in all aspects of social life. The capital market (and companies) is at the forefront of these phenomena, given its volatility and extreme exposure to these macro events.
In this scenario, the objective was to develop a platform that predicts insolvencies. The Riskit: Insolvency
Predictor is a web-based platform aimed at assisting the scientific community and investors in
predicting the possibility of companies becoming insolvent based on specific financial indicators.
Methodologically, a dataset of 15,000 Portuguese companies was randomly extracted from the Iberian
Balance Sheet Analysis System (SABI) database1. An analysis was conducted, resulting in the selection
of 11 financial indicators used for predictions. To make predictions, the authors conducted a
comprehensive study of models commonly used for this type of forecasting and also experimented
with some machine-learning models that are not frequently mentioned in the literature. The evaluation
of the application’s performance in predicting insolvencies is measured by a series of performance
benchmarks calculated with the help of a confusion matrix.
It was found that models frequently mentioned in the literature do not always have better performance.
The main objectives of this project were achieved, providing both the scientific community
and investors with a tool that predicts insolvency using a set of financial indicators and demonstrating
the value of machine-learning models for making these predictions. The application can be visited at
https://riskit.ipb.pt/.
Description
Mestrado em IPB-ESTG e ASSOCIAÇÃO DE POLITÉCNICOS DO NORTE (APNOR): Instituto Politécnico do Cávado e do Ave, P. Porto, Instituto Politécnico de Viana do Castelo
Keywords
Insolvency prediction Risk management Financial indicators Machine learning models Web-based application