Researcher ORCID Identifier
Date of Submission
Campus Only Senior Thesis
Bachelor of Arts
© 2021 Tumisang M Mosedame
A typical successful technology company can have good or poor accounting ratios and still have a high market value. There seems to be a non-linearity in the relationship between accounting ratios and market values; we can thus assume a non-linear relationship between accounting ratios and bankruptcy. We are yet to study how this affects traditional bankruptcy prediction models, which often require a linear relationship between the independent and dependent variables. This study will contrast traditional statistical methods with more advanced methods of bankruptcy prediction in information technology companies(IT) from GICS IT sector code, 45. I randomly partition my data into half using 50 percent of it for test data and the other half for training data to develop or build the models. The models are Logistic Regression, Linear Discriminant Approaches, Decision tree classification, Bagged decision trees, Naive Bayes, and Perceptron. In each of these models, I evaluate each of the models for accuracy, specificity, sensitivity, precision, and error rate to determine the best model. I find that the best model for bankruptcy prediction for IT companies is the Decision Tree Model as it had the best scores in four of our five evaluative measures. It scored 96.90 in accuracy, 99.40 in sensitivity, 96.40 in the precision measure, and had the second smallest error measure at 3.26 percent.
Mosedame, Tumisang, "Predicting Bankruptcy of Information Technology Companies: a comparison of alternative methods" (2021). CMC Senior Theses. 2779.
This thesis is restricted to the Claremont Colleges current faculty, students, and staff.