Medical Data Classification Using Binary Brain Storm Optimization

The volume of data in the medical domain has been on the rise with improved and accessible technologies; this big size of data increased the complexity of the process of analysis and knowl- edge discovery and thus making it more difficult to match patterns. Hence, it is necessary to use feature selection and classification models on medical diagnosis because it reduces the complexity of the data volume by using the non-trivial features in leading to more guided medical research. Binary Brain Storming Optimization (BBSO) is one of the many heuristic algorithms that have been applied on non-medical data with good reported performance. This work is a study to de- termine the performance of the BBSO on medical data through comparative analysis with other feature selecting algorithms using different classifiers. In this work, BBSO, along with three other feature selecting algorithms, namely: Binary Particle Swarm Optimization, Binary Grey Wolf Optimization, and the Genetic Algorithms, which had previously been applied in medical classifi- cation problems, were applied on five different medical data. Using Feature Selection algorithms, redundant, noisy, and irrelevant attributes are removed reducing the complexity of: the data, the classifiers model and the computational time. The resulting subset usually leads to a better clas- sifier performance. Their resulting features were utilized to develop classifier models within six different classifiers: K-Nearest Neighbors, Decision Tree, Na ̈ıve-Bayes, Random Forest, Linear Discriminant analysis and four different hyper-parameter variants of the Support Vector Machine. The results from the research demonstrated good performance on medical data, making BBSO a good Feature Selection algorithm for medical diagnosis.