B1494
Title: Feature ranking in the diagnosis of cardiovascular diseases
Authors: Paria Sarzaeim - Alzahra University (Iran)
Arash Negahdari Kia - University of Limerick (Ireland) [presenting]
Kevin Burke - University of Limerick (Ireland)
Parisa Shamsi - Alzahra University (Iran)
Abstract: Since cardiovascular diseases such as heart attacks are one of the leading causes of death annually, it is essential to diagnose them correctly or more accurately. Many reasons may cause cardiovascular diseases, like high blood pressure or high cholesterol. Such symptoms can help diagnose cardiovascular diseases. Developing prediction models requires an understanding of which predictors are most relevant in diagnosis. There are many machine learning algorithms for feature selection, each with advantages and disadvantages. The first step is to do an exhaustive search on a range of machine learning predictors. This will enable us to find the models with the highest performance using different evaluation criteria. The effect of each feature elimination on various models and evaluation criteria is calculated in best-fit models. These reductions are then used in a novel weighted average score to calculate the importance score of each feature. Cleveland and Kaggle datasets for Heart Disease Diagnosis are employed. Finally, features are ranked based on their score and the results regarding the best-fit models and feature ranks are discussed.
