A Hybrid Machine Learning Model for Credit Approval

Figures & data

Figure 1. The proposed framework.

Table 1. Comparison of works

Work	Techniques	Evaluation
Chen and Li (2010)	Feature selection (FS) + Classifiers (SVM)	Accuracy and error rates
Hsieh (2005)	Clustering (KM) + Classification (ANN)	Accuracy and error rates
Hsieh and Hung (2010)	Clustering (KM) + Classification (ANN)	Association rules for bad credit status
Huang, Chen, and Wang (2007)	Feature selection (FS) + Classifiers (SVM)	Accuracy rates
Huysmans et al. (2006)	Classification + Clustering (Self-Organizing Map, SOM)	Accuracy rates
Koutanaei, Sajedi, and Khanbabaei (2015)	Feature selection (FS) + Classifiers	Accuracy and error rates
Lee et al. (2002)	Classification + Classification	Accuracy and error rates
Luo, Cheng, and Hsieh (2009)	Classifiers	Accuracy rates
Malhotra and Malhotra (2002)	Classification + Classification	Accuracy rates
Ping and Yongheng (2011)	Feature selection (FS) + Classifiers (SVM)	Accuracy rates
Zhao et al. (2015)	Instance selection (IS) + Classifiers (Multilayer perceptron, MLP)	Error rates
Pristyanto, Adi, and Sunyoto (2019)	Feature selection (FS) + Classifiers	Accuracy rates

Figure 2. The CBDT algorithm.

Figure 3. EM_Clustering Subroutine.

Figure 4. Generate_Decision_Tree Subroutine.

Table 2. Comparison of the six approaches

Approaches	Features selection	Instances selection	Prediction model
DT	None	None	C4.5
MLP	None	None	Multiple Layer perceptron
NB	None	None	Naive Bayes classifier
RF	None	None	Random Forest
SVM	None	None	Support Vector Machine
CBDT (this study)	Gain-ratio	EM	C4.5

Table 3. Parameter settings of the proposed CBDT model

Aim	Model	Parameters
Clustering	EM	Cluster_Count = 2
		Clustering_Method = 1 (EM)
		Mininmum_support = default
		Modeling_Cardinality = default
		Stoping_Tolerance = default
Prediction	C4.5	Complexity_Penalty = default
		Mininmum_support = default
		Score_Method = default
		Split_Method = default

Table 4. Confusion matrix for positive and negative tuples

	Predicted Class
Actual Class		C1 (P)	C2 (N)
	C1 (P)	True Positives (TP)	False Negatives (FN)
	C2 (N)	False Positives (FP)	True Negatives (TN)

Table 5. Cost matrix

		Predicted Class
		P	N
Actual Class	P	TPC	FNC
	N	FPC	TNC

Table 6. Cost matrix for credit approval

		Predict
		P(Good)	N(Bad)
Actual	P(Good)	0	1
	N(Bad)	5	0

Table 7. The average experiment results compared with the other 5 methods

Method	Precision	TPR/Recall	F1	Accuracy	CostEffect
CBDT	0.93	0.91	0.92	0.89	0.28
DT	0.92	0.84	0.88	0.84	0.35
MLP	0.74	0.96	0.84	0.73	1.24
NB	0.84	0.80	0.82	0.75	0.69
RF	0.79	0.94	0.86	0.78	0.91
SVM	0.85	0.94	0.89	0.83	0.67

Table 8. The average experiment results of the new three methods

Method	Precision	TPR	F1	Accuracy	CostEffect
CBMLP (MLP)	0.74 (0.74)	0.97 (0.96)	0.84 (0.84)	0.73 (0.73)	1.24 (1.24)
CBNB (NB)	0.80 (0.84)	0.39 (0.80)	0.53 (0.82)	0.50 (0.75)	0.78 (0.69)
CBSVM (SVM)	0.86 (0.85)	0.96 (0.94)	0.91 (0.89)	0.86 (0.83)	0.58 (0.67)

Table 9. The average experiment results compared with the other three methods

Method	Precision	TPR	F1	Accuracy	CostEffect
CBDT	0.93	0.91	0.92	0.89	0.28
CBMLP	0.74	0.97	0.84	0.73	1.24
CBNB	0.80	0.39	0.53	0.50	0.78
CBSVM	0.86	0.96	0.91	0.86	0.58

Table 10. The average experiment results compared with the other five methods

Method	Precision	TPR/Recall	F1	Accuracy	CostEffect
CBDT	0.92	0.92	0.92	0.93	0.20
DT	0.81	0.90	0.85	0.85	0.56
MLP	0.76	0.51	0.46	0.60	1.18
NB	0.82	0.62	0.70	0.78	0.46
RF	0.76	0.87	0.80	0.79	0.83
SVM	0.64	0.67	0.57	0.71	0.87

Table 11. The average experiment results of the new three methods

Method	Precision	TPR	F1	Accuracy	CostEffect
CBMLP (MLP)	0.68 (0.76)	0.93 (0.51)	0.78 (0.46)	0.78 (0.60)	0.99 (1.18)
CBNB (NB)	0.85 (0.82)	0.61 (0.62)	0.71 (0.70)	0.78 (0.78)	0.41 (0.46)
CBSVM (SVM)	0.92 (0.64)	0.96 (0.67)	0.94 (0.57)	0.94 (0.71)	0.22 (0.87)

Table 12. The average experiment results compared with the other three methods

Method	Precision	TPR	F1	Accuracy	CostEffect
CBDT	0.92	0.92	0.92	0.93	0.20
CBMLP	0.68	0.93	0.78	0.78	0.99
CBNB	0.85	0.61	0.71	0.78	0.41
CBSVM	0.92	0.96	0.94	0.94	0.22

Table 13. The average experiment results compared with the other five methods

Method	Precision	TPR/Recall	F1	Accuracy	CostEffect
CBDT	0.95	0.91	0.93	0.94	0.16
DT	0.93	0.94	0.93	0.94	0.20
MLP	0.71	0.56	0.62	0.70	0.72
NB	0.86	0.66	0.75	0.80	0.39
RF	0.90	0.88	0.89	0.90	0.28
SVM	0.72	0.67	0.67	0.84	0.19

Table 14. The average experiment results of the new three methods

Method	Precision	TPR	F1	Accuracy	CostEffect
CBMLP (MLP)	0.80 (0.71)	0.82 (0.56)	0.78 (0.62)	0.80 (0.70)	0.66 (0.72)
CBNB (NB)	0.75 (0.86)	0.57 (0.66)	0.64 (0.75)	0.69 (0.80)	0.76 (0.39)
CBSVM (SVM)	0.92 (0.72)	0.65 (0.67)	0.67 (0.67)	0.84 (0.84)	0.17 (0.19)

Table 15. The average experiment results compared with the other three methods

Method	Precision	TPR	F1	Accuracy	CostEffect
CBDT	0.95	0.91	0.93	0.94	0.16
CBMLP	0.80	0.82	0.78	0.80	0.66
CBNB	0.75	0.57	0.64	0.69	0.76
CBSVM	0.92	0.65	0.67	0.84	0.17

Chen, F. L., and F. C. Li. 2010. Combination of feature selection approaches with SVM in credit scoring. Expert Systems with Applications 37 (7):4902–09. doi:https://doi.org/10.1016/j.eswa.2009.12.025.

 Web of Science ®Google Scholar

Hsieh, N. C. 2005. Hybrid mining approach in the design of credit scoring models. Expert Systems with Applications 28 (4):655–65. doi:https://doi.org/10.1016/j.eswa.2004.12.022.

 Web of Science ®Google Scholar

Hsieh, N. C., and L. P. Hung. 2010. A data driven ensemble classifier for credit scoring analysis. Expert Systems with Applications 37 (1):534–45. doi:https://doi.org/10.1016/j.eswa.2009.05.059.

 Web of Science ®Google Scholar

Huang, C. L., M. C. Chen, and C. J. Wang. 2007. Credit scoring with a data mining approach based on support vector machines. Expert Systems with Applications 33 (4):847–56. doi:https://doi.org/10.1016/j.eswa.2006.07.007.

 Web of Science ®Google Scholar

Huysmans, J., B. Baesens, J. Vanthienen, and T. Van Gestel. 2006. Failure prediction with self-organizing maps. Expert Systems with Applications 30 (3):479–87. doi:https://doi.org/10.1016/j.eswa.2005.10.005.

 Web of Science ®Google Scholar

Koutanaei, F. N., H. Sajedi, and M. Khanbabaei. 2015. A hybrid data mining model of feature selection algorithms and ensemble learning classifiers for credit scoring. Journal of Retailing and Consumer Services 27:11–23. doi:https://doi.org/10.1016/j.jretconser.2015.07.003.

 Web of Science ®Google Scholar

Lee, T. S., C. C. Chiu, C. J. Lu, and I. F. Chen. 2002. Credit scoring using the hybrid neural discriminant technique. Expert Systems with Applications 23 (3):245–54. doi:https://doi.org/10.1016/S0957-4174(02)00044-1.

 Web of Science ®Google Scholar

Luo, S. T., B. W. Cheng, and C. H. Hsieh. 2009. Prediction model building with clustering-launched classification and support vector machines in credit scoring. Expert Systems with Applications 36 (4):7562–66. doi:https://doi.org/10.1016/j.eswa.2008.09.028.

 Web of Science ®Google Scholar

Malhotra, R., and D. K. Malhotra. 2002. Differentiating between good credits and bad credits using neuro-fuzzy systems. European Journal of Operational Research 136 (1):190–211. doi:https://doi.org/10.1016/S0377-2217(01)00052-2.

 Web of Science ®Google Scholar

Ping, Y., and L. Yongheng. 2011. Neighborhood rough set and SVM based hybrid credit scoring classifier. Expert Systems with Applications 38 (9):11300–04. doi:https://doi.org/10.1016/j.eswa.2011.02.179.

 Web of Science ®Google Scholar

Zhao, Z., S. Xu, B. H. Kang, M. M. J. Kabir, Y. Liu, and R. Wasinger. 2015. Investigation and improvement of multi-layer perceptron neural networks for credit scoring. Expert Systems with Applications 42 (7):3508–16. doi:https://doi.org/10.1016/j.eswa.2014.12.006.

 Web of Science ®Google Scholar

Pristyanto, Y., S. Adi, and A. Sunyoto (2019, July). The effect of feature selection on classification algorithms in credit approval. In 2019 International Conference on Information and Communications Technology (ICOIACT) (pp. 451–56). IEEE, Yogyakarta, Indonesia.

 Google Scholar