Open AccessA Comparison of Feature Selection Methodology for Solving Classification Problems in Finance
Author(s) -
Xia Xiaomao,
Xudong Zhang,
YuanFang Wang
Publication year - 2019
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1284/1/012026
Subject(s) - feature selection , feature (linguistics) , computer science , selection (genetic algorithm) , artificial intelligence , shapley value , machine learning , value (mathematics) , set (abstract data type) , data mining , pattern recognition (psychology) , mathematics , game theory , philosophy , linguistics , mathematical economics , programming language
Classification is one of the most important tasks in real world with the intention of finding the underlying patterns of the data and making use of the found patterns. Other than considering lots of parameters that may influence the classification model accuracy, the influence of the feature selections has been rarely discussed. SHAP (SHapley Additive exPlanation) value is a novel ensemble learning measurement that is the unique consistent and locally accurate attribution value. Its applications to feature selection are not mentioned in the literature. In this paper, we compare the prediction accuracy of several feature selection techniques on a set of Polish companies. We show that new feature selection method based on the SHAP values is superior (not worse) to other widely used prediction methods.