An Ensemble Feature Selection Methodology That Incorporates Domain Knowledge for Cardiovascular Disease Diagnosis


KOLUKISA B., GÜNGÖR V. Ç., Gungor B. B.

28th Signal Processing and Communications Applications Conference (SIU), ELECTR NETWORK, 5 - 07 Ekim 2020 identifier identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/siu49456.2020.9302048
  • Basıldığı Ülke: ELECTR NETWORK
  • Anahtar Kelimeler: Data mining, Machine Learning, Classification Algorithm, Coronary Artery Disease, HEART-DISEASE
  • Abdullah Gül Üniversitesi Adresli: Evet

Özet

Coronary Artery Disease (CAD) is the condition where, the heart is not fed enough as a result of the accumulation of fatty matter called atheroma in the walls of the arteries. In 2016, CAD accounts for 31% (17.9 million) of the world's total deaths and its diagnosis is difficult. It is estimated that approximately 23.6 million people will die from this disease in 2030. With the development of machine learning and data mining techniques, it might be possible to diagnose CAD inexpensively and easily via examining some physical and biochemical values. In this study, for the CAD classification problem, a novel ensemble feature selection methodology that incorporates domain knowledge is proposed. Via applying the proposed methodology on the UCI Cleveland CAD dataset and using different classification algorithms, performance metrics are compared. It is shown that in our experiments, when Multilayer Perceptron classifier is used with 9 selected features, our proposed solution reached 85.47% accuracy, 82.96% accuracy and 0.839 F-Measure. As a future work, we aim to generate a machine learning model that can quickly diagnose CAD on real-time data in hospitals.