Adaptive Softmax Regression for Credit Scoring. Munkhdalai, L., Davagdorj, K., Pham, V. H., & Ryu, K. H. Smart Innovation, Systems and Technologies, 212:409–417, 2021. ISBN: 9789813367562
doi  abstract   bibtex   
Credit scoring is a classification task from the machine learning perspective. Efficiently classifying bad borrowers is the main aim of building a credit scoring model. This work proposes a novel adaptive softmax regression method for credit scoring. We augment a simple softmax regression by deep neural networks to make its estimated probabilities as an adaptive for each observation. Our experimental result on public benchmark datasets shows that adaptive softmax regression outperformed the machine learning baselines in terms of Brier score, area under the curve (AUC) and accuracy.
@article{Pham2021,
	title = {Adaptive {Softmax} {Regression} for {Credit} {Scoring}},
	volume = {212},
	issn = {21903026},
	doi = {10.1007/978-981-33-6757-9_51},
	abstract = {Credit scoring is a classification task from the machine learning perspective. Efficiently classifying bad borrowers is the main aim of building a credit scoring model. This work proposes a novel adaptive softmax regression method for credit scoring. We augment a simple softmax regression by deep neural networks to make its estimated probabilities as an adaptive for each observation. Our experimental result on public benchmark datasets shows that adaptive softmax regression outperformed the machine learning baselines in terms of Brier score, area under the curve (AUC) and accuracy.},
	journal = {Smart Innovation, Systems and Technologies},
	author = {Munkhdalai, Lkhagvadorj and Davagdorj, Khishigsuren and Pham, Van Huy and Ryu, Keun Ho},
	year = {2021},
	note = {ISBN: 9789813367562},
	keywords = {Adaptive learning, Credit scoring, Decision making, Softmax regression},
	pages = {409--417},
}

Downloads: 0