Developing an Intelligent System Based on Knowledge Engineering Paradigms for Bankruptcy Prediction

Abstract

Constructing a robust machine learning system without classification errors is a very important challenge in predicting bankruptcy. Since misclassification errors are due to biased distribution between classes. Initially, this research clearly illustrated the significant contributions of machine learning in predicting bankruptcy. This research presented variety of machine learning techniques with different aspects. Moreover, it provided different re-sampling techniques to differentiate between them. It showed that oversampling techniques are more efficient than under-sampling techniques. This research developed an efficient machine learning system to predict bankruptcy. The developed system re-sampled the used dataset to achieve a better performance. Moreover, it aimed to remove the redundant features for more accurate classification results. The developed system was evaluated on different public datasets from UCI repository.