CLASSIFICATION AND FEATURE SELECTION IN MEDICAL DATA PREPROCESSING
Akhram Khasanovich Nishanov, G.P. Djurayev, Malika Khasanova · Zenodo (CERN European Organization for Nuclear Research) · 2020
In this article, the issues such as medical data preprocessing, reclassification of training sets and determining the importance of classes, forming reference tables were solved. As a result of data preprocessing, three objects were formed: 1) Ischemic heart disease. Unstable angina pectoris; 2) Ischemic heart disease. Acute myocardial infarction; 3) Ischemic heart disease. Arrhythmic form. Further, the issues such as classifying, selecting a set of informative features that differentiate between class objects were solved using Fisher criterion and algorithms for an estimated calculation as well as software programs for them were developed. As a result data preprocessing reference classes were formed. Objects that had fallen outside of their class during the formation process were excluded from the training set. A classification and a set of informative features were selected using established classes. Initially, three class objects each containing 62 features were provided by medical professionals and as a result data preprocessing three sets consisting of 131, 115, and 40 objects respectively in three classes were used to form a reference table.