검색 상세

(THREE ESSAYS ON) KNOWLEDGE CONSOLIDATION APPROACHES FOR FINANCIAL DECISION MAKING

초록/요약

Bankruptcy prediction and credit scoring have long been regarded as critical topics and have been studied extensively in the accounting and finance literature. The main impacts of such research are in lending decisions and profitability of financial institutions. Before extending a loan, banks need to predict the possibility of failure of the potential counterparty. Thus, predicting bankruptcy timely and correctly has become great importance for financial institutions. Further, the benefits of using credit scoring include reducing the cost of credit analysis, enabling faster decision, insuring credit collections, and diminishing possible risk. A slight improvement in credit scoring accuracy might reduce large credit risk and translate into significant future saving. Artificial intelligence and machine learning techniques have been used to solve these financial decision-making problems. Neural networks, especially, have received a lot of attention because of their universal approximation property. However, a major drawback associated with the use of neural networks for decision making is their lack of explanation capability. This dissertation proposes three essays on the knowledge consolidation approaches to solve financial decision-making problems. The architecture of knowledge consolidation approaches is also called an ensemble approach. The key idea in knowledge consolidation approaches is to combine a number of classifiers such that the resulting combined system achieves higher classification accuracy and efficiency than the original single classifiers. The aim of knowledge consolidation approaches is to design a composite system that outperforms any individual classifier by pooling together the decisions of all classifiers. Compared to single classifiers, knowledge consolidation approaches is better in its performance as it consolidates knowledge from single classifiers. Another advantage of knowledge consolidation approaches is that it does not need the memory space to store the dataset as only extracted knowledge is necessary in build this integrated model. It can also reduce the costs from storage allocation, memory, and time schedule. The first approach investigates to develop a knowledge integration (KI) model for the prediction of future dividends. The effectiveness of KI model was verified by the experiments comparing with Marsh & Merton (M&M), neural networks, and CART approach. The second approach investigates that the feasibility and effectiveness of a knowledge consolidation model (KCM) for the credit rating evaluation. The KCM combines the rules extracted using KBANN (NeuroRule), Frequency Matrix (which is similar to the Na?ve Bayesian technique), and C5.0 algorithm. The results from the tests show that the performance of KCM is superior to that of the other single models such as multiple discriminant analysis, logistic regression, frequency matrix, neural networks, decision trees, and NeuroRule. The last approach investigates that the feasibility and effectiveness of a rule accumulation algorithm (RAA) and a GA-based rule refinement algorithm (GA-RRA) for the credit rating evaluation. The experiment compares the performance of the random dataset, RAA, elimination of redundant rules (ERR), and GA-RRA. The results from the tests show that the performance of the GA-RRA is superior to that of the other algorithms such as random dataset, ERR, and RAA.

more