Respiratory Disease Pre-Diagnosis through a Novel Pattern Classification Algorithm based on Associative Memories

In this paper, the Subtractive Threshold Associative Classifier (STAC), a novel supervised machine learning model, is presented. The main contribution of the proposed model is to have the capability to adequately deal with medical dataset for the pre-diagnosis of respiratory disease and class imbalance data complexity without applying any other pre-processing technique, obtained competitive results. Furthermore, the proposed algorithm is interpretable and transparent, since the reasons why a test pattern was classified as belonging to a specific class. The experimental results were validated with the purpose of finding possible significant differences in performance; For this, statistical tests were used. It is necessary to emphasize that the experimental tests carried out allow us to verify that the new proposed algorithm is competitive against the most used algorithms in the state of the art.