Akademik Veri Yönetim Sistemi
14th Turkish Congress of Medical Informatics Association, İzmir, Türkiye, 16 - 18 Mart 2023, ss.38-44
Advances in medicine enable the production of large amounts of data
and the discovery of alternative treatment methods using this data, and valuable
information. By using data mining and machine learning methods, it is possible
to analyze big data and extract important information from these data. The
biggest challenge in information extraction is the presence of many unnecessary
variables that reduce the performance of machine learning methods and
disadvantage in terms of computation time and complexity. In this study, binary
version of the PSO-GWO optimization algorithm, which is a hybrid
optimization method, is applied to medical data to extract these unnecessary
variables from the model and the accuracy comparisons are made by using
different classifiers. Two data sets are used for comparisons. In both data sets,
instead of using all the variables, classification using the variables found with
the hybrid approach gives higher accuracy values. As a result, using a reduced
number of features instead of all features increases the performance of the
classifier, avoiding the use of many unnecessary variables, provides advantages
in terms of computation time.