Implementing binary particle swarm optimization and C4.5 decision tree for cancer detection based on microarray data classification

Cancer is one of deadly disease in the world and needed to detect the symptoms early. Cancer can be represented with microarray data with measuring the changes occured in gene expression level. Cancer detection can be done by doing classification technique for microarray data. One of most algorithm that applied for classification is C4.5 Decision Tree. It is a linier method which is easy to interpret and included into the algorithm which has given impact in classification but it is sensitive to noise data. Microarray data has a large features (high dimensional) which is not all the features has important information (high noise) and small samples which is causing the classification is difficult and affect the accuracy. Binary Particle Swarm Optimization (BPSO) is one of search optimization algorithm that could find the optimal feature. The purpose in this research consists of implementing and analysing the influence of feature selection and classification on microarray data using Binary Particle Swarm Optimization (BPSO) as feature selection and Decision Tree C4.5 as classifier. The discretization is needed for Decision Tree rule model and applied using K-Means. System is divided into two schemes such as Information Gain (IG) – C4.5 and BPSO – C4.5. The accuracy result based on IG – C4.5 and BPSO – C4.5 both are 54% and 99%. Applying feature selection before the classification could avoid the noise data in microarray data so it could form the rule accurately. With applying BPSO and Decision Tree is able to find the most significant feature and improve the accuracy.