Decision diagrams in machine learning: an empirical study on real-life credit-risk data

One of the key decisions financial institutions have to make as part of their daily operations is to decide whether or not to grant a loan to an applicant. With the emergence of large-scale data-storing facilities, huge amounts of data have been stored regarding the repayment behavior of past applicants. It is the aim of credit scoring to analyze this data and build models that distinguish good from bad payers using characteristics such as amount on savings account, marital status, purpose of loan, etc. Many classification techniques have been suggested to build credit-scoring models. Especially neural networks have in recent years received a lot of attention. However, while they are generally able to achieve a high predictive accuracy rate, the reasoning behind how they reach their decisions is not readily available, which hinders their acceptance by practitioners. Therefore, in [1], we have proposed a two-step process to open the neural network black box which involves: (1) extracting rules from the network; (2) visualizing this rule set using an intuitive graphical representation, such as decision tables or trees.

[1]  A. Steenackers,et al.  A credit scoring model for personal loans , 1989 .

[2]  Emden R. Gansner,et al.  A Technique for Drawing Directed Graphs , 1993, IEEE Trans. Software Eng..

[3]  Usama M. Fayyad,et al.  Multi-Interval Discretization of Continuous-Valued Attributes for Classification Learning , 1993, IJCAI.

[4]  Jude W. Shavlik,et al.  in Advances in Neural Information Processing , 1996 .

[5]  Douglas Hall,et al.  A new functional decomposition method as applied to machine learning and vlsi layout , 2000 .

[6]  Masahiro Fujita,et al.  On variable ordering of binary decision diagrams for the application of multi-level logic synthesis , 1991, Proceedings of the European Conference on Design Automation..

[7]  Ron Kohavi,et al.  Oblivious Decision Trees, Graphs, and Top-Down Pruning , 1995, IJCAI.

[8]  Alexander Gammerman,et al.  Machine-learning algorithms for credit-card applications , 1992 .

[9]  Kenneth J. Supowit,et al.  Finding the Optimal Variable Ordering for Binary Decision Diagrams , 1990, IEEE Trans. Computers.

[10]  Vijay S. Desai,et al.  A comparison of neural networks and linear scoring models in the credit union environment , 1996 .

[11]  L. Thomas.  A survey of credit and behavioural scoring: forecasting financial risk of lending to consumers , 2000 .

[12]  David J. Hand,et al.  Construction of a k-nearest-neighbour credit-scoring system , 1997 .

[13]  Johan A. K. Suykens,et al.  Benchmarking state-of-the-art classification algorithms for credit scoring , 2003, J. Oper. Res. Soc..

[14]  Randal E. Bryant,et al.  Graph-Based Algorithms for Boolean Function Manipulation , 1986, IEEE Transactions on Computers.

[15]  J. Ross Quinlan,et al.  C4.5: Programs for Machine Learning , 1992 .

[16]  Huan Liu,et al.  Symbolic Representation of Neural Networks , 1996, Computer.

[17]  Ron Kohavi,et al.  Wrappers for performance enhancement and oblivious decision graphs , 1995 .

[18]  Bart Baesens,et al.  Using Neural Network Rule Extraction and Decision Tables for Credit - Risk Evaluation , 2003, Manag. Sci..

[19]  Toshihide Ibaraki,et al.  Ordered binary decision diagrams as knowledge-bases , 2002, Artif. Intell..

[20]  Tiziano Villa,et al.  Multi-valued decision diagrams: theory and applications , 1998 .

[21]  Tapio Elomaa,et al.  On the Practice of Branching Program Boosting , 2001, ECML.

[22]  Arlindo L. Oliveira,et al.  Using the Minimum Description Length Principle to Infer Reduced Ordered Decision Graphs , 1996, Machine Learning.