THE 3
rd
INTERNATIONAL SCIENTIFIC CONFERENCES OF STUDENTS AND YOUNG RESEARCHERS
dedicated to the 99
th
anniversary of the National Leader of Azerbaijan Heydar Aliyev
125
classifiers used for credit risk evaluation can be categorised into two families:
Supervised Learning Methods: Two mostly used single classifiers for this
task are Support Vector Machines (SVM), Random Forest and Neural
Networks. Unsupervised Learning Methods: Clustering methods can also be
used to identify the risk of bankruptcy or credit default. These methods can
help to identify groups of loan applicants/ enterprises with similar
characteristics. A cluster-based dynamic scoring model can achieve better
scoring accuracy by implementing different classifiers for different clusters.
In this paper, core machine learning models, namely SVC, random forest,
and NNs with deep learning were employed, and the performance of all
models was compared. (ROC is a probability curve and AUC represents the
degree or measure of separability. It tells how much the model is capable of
distinguishing between classes.)
The ROC AUC score of SVC
for
first cluster
0.5000
The ROC AUC score of RF
for
first cluster
0.5906
The ROC AUC score of NN
for
first cluster
0.6155
Dostları ilə paylaş: