Authors:
(1) Harrison Mateika, Northwestern University (HarrisonMateika2022@u.northwestern.edu);
(2) Juannan Jia, Northwestern University (JuannanJia2020@u.northwestern.edu);
(3) Linda Lillard, Northwestern University (LindaLillard2022@u.northwestern.edu);
(4) Noah Cronbaugh, Northwestern University (NoahCronbaugh2022@u.northwestern.edu);
(5) Will Shin, Northwestern University (WilliamShin2022@u.northwestern.edu). Authors: Authors: (1) Harrison Mateika, Northwestern University (HarrisonMateika2022@u.northwestern.edu); (2) Juannan Jia, Northwestern University (JuannanJia2020@u.northwestern.edu); (3) Linda Lillard, Northwestern University (LindaLillard2022@u.northwestern.edu); (4) Noah Cronbaugh, Northwestern University (NoahCronbaugh2022@u.northwestern.edu); (5) Will Shin, Northwestern University (WilliamShin2022@u.northwestern.edu). Table of Links Introduction
Literature Review
Data Collection
Data Analysis
Methodology
Results
Analysis and Interpretation
Conclusions and Next Steps, and References Introduction Introduction Literature Review Literature Review Data Collection Data Collection Data Analysis Data Analysis Methodology Methodology Results Results Analysis and Interpretation Analysis and Interpretation Conclusions and Next Steps, and References Conclusions and Next Steps, and References The primary aim of this research was to find a model that best predicts which fallen angel bonds would either potentially rise up back to investment grade bonds and which ones would fall into bankruptcy. To implement the solution, we thought that the ideal method would be to create an optimal machine learning model that could predict bankruptcies. Among the many machine learning models out there we decided to pick four classification methods: logistic regression, k-nearest neighbor, support vector machines, and neural networks. We also utilized an automated methods of machine learning: Google Cloud’s AutoML. The results of our model comparisons showed that the models did not predict bankruptcies very well on the original data set with the exception of AutoML having a high precision score. However, our oversampled and feature selection data set did perform very well. However, this could likely be due to the model being overfitted to match the narrative of the oversampled data (as in, it doesn’t accurately predict data outside of this data set quite well). Therefore, we were not able to create a model that we are confident that would predict bankruptcies. However, we were able to find value out of this project in two key ways. The first is that the AutoML model in every metric and in every data set either outperformed or performed on par with the other models. The second is that we found that utilizing feature selection did not reduce predictive power that much. This means that we can reduce the amount of data to collect for future experimentation regarding predicting bankruptcies. 1. Introduction The goal of investing in bonds is to maximize the return on an investment to the highest degree possible. This can often mean the blending together of two diametrically opposed methods: investing in conservative bonds that are guaranteed a return, or investing in aggressive bonds that carry a substantially increased amount of risk. The conservative bonds tend to guarantee a return, but the return is fairly low due to its low yield (also known as the interest an investor receives on a bond). However, the riskier bonds have a higher yield and thus can result in a far greater return. Due to this, investors often have to make the unenviable decision to invest in bonds that they know are at a higher risk of failure. The bond market is in a unique position when compared to other financial markets due to its rating system. This rating system is oligarchical by nature, since the ratings are given out by three bond rating agencies. All of these agencies follow the exact same ratings system: AAA is the highest, while D is the lowest. Bonds that are highly rated, and thus conservative, are referred to as investment-grade bonds. These bonds typically fall in the AAA, AA, A, or BBB ratings. Bonds that are rated lower than that are referred to as non-investment-grade bonds. These bonds are also unceremoniously known as junk bonds. Junk bonds usually consist of two types of companies: small startups or companies with high debt ratios. These are normally the types of risky bonds that investors struggle to invest in. For this reason, they are often looking for an opportunity to mitigate the risk of failure. One potential opportunity is with Fallen Angel and Rising Star bonds. A Fallen Angel refers to a bond that once had an investment-grade rating, but fell to junk bond territory due to its company’s financial hardship. A Rising Star refers to a bond that was once considered a junk bond, but rose to an investment-grade bond. The essential idea is that if an investor invests in a Fallen Angel, the investor can take advantage of the bond’s currently high yield and low purchase price. If the bond were to become a Rising Star, the bond’s price would rise, and allow the investor to sell it for a profit. However, the investor also is taking on a major risk. Since a Fallen Angel is often a company that has fallen on hard times, it also is likely that the bond could continue to fall into bankruptcy. Fallen Angels and junk bonds are everywhere and often one of the challenges associated with them is understanding the inherent risk of investing in these bonds. While the rating system is a helpful barometer, the amount of information it gives regarding whether or not a company could go bankrupt is dubious. Further complicating the matter is the amount of information that goes into determining how a company is performing and what direction it’s going in. The amount of information is far too much for an individual to take in, and thus requires methods beyond financial ratios and eye-scans. The challenge of determining the chance of a company going bankrupt, therefore, deserves a degree of scrutiny that cannot be accomplished using typical methods. We would like to find the optimal machine learning method that can best predict whether or not a company will go bankrupt by running various models manually, and then comparing them not only against each other but also against an automated machine learning tool called AutoML, a product on Google Cloud. This program sets out to implement feature engineering, dimensional reduction, and model choice to ease the process of machine learning. This paper is available on arxiv under CC by 4.0 Deed (Attribution 4.0 International) license. This paper is available on arxiv under CC by 4.0 Deed (Attribution 4.0 International) license. available on arxiv

Part of HackerNoon's growing list of open-source research papers, promoting free access to academic material.

Can Machines Spot Rising Stars? Exploring AI‑Driven Insight into Fallen Angel Bonds

About Author

Comments

TOPICS

THIS ARTICLE WAS FEATURED IN

Related Stories

Addressing Data Imbalance in Bankruptcy Prediction: Comparative Literature Insights

Automated Machine Learning for Data Analysts & Business Users

Enabling Scientific Research and Analyses Through Automated ML

Machine Learning Meets HR: Predicting Employee Attrition with PyCaret

Addressing Data Imbalance in Bankruptcy Prediction: Comparative Literature Insights

Data Collection and Analysis for Machine Learning‑Based Bankruptcy Prediction

Addressing Data Imbalance in Bankruptcy Prediction: Comparative Literature Insights

Automated Machine Learning for Data Analysts & Business Users

Enabling Scientific Research and Analyses Through Automated ML

Machine Learning Meets HR: Predicting Employee Attrition with PyCaret

Addressing Data Imbalance in Bankruptcy Prediction: Comparative Literature Insights

Data Collection and Analysis for Machine Learning‑Based Bankruptcy Prediction

Light-Mode

Classic

Newspaper

Minty

Dark-Mode

Neon Noir

Minty

HN StartUps