EMOPS: An Enhanced Multi-Objective Particle Swarm Based Classifier for Poorly Understood Cancer Patterns

摘要

Microarray based cancer pattern classification is one of the popular techniques in bioinformatics research. At the same time, it was noticed that for studying the expression levels through the gene expression profiling experiments, thousands of genes have to be simultaneously studied to understand the patterns of the gene expression or cancer pattern. This research proposed an efficient cancer pattern classifier called an Enhanced Multi-Objective Particle Swarm (EMOPS) and it is studied thoroughly in terms of memory utilization, execution time (processing time), sensitivity, specificity, classification accuracy and F-score. The results were compared with that of the recently proposed classifiers namely Hybrid Ant Bee Algorithm (HABA), Kernelized Fuzzy Rough Set based semi supervised Support Vector Machine (KFRS-S3VM) and Multi-objective Particle Swarm Optimization (MPSO). For analyzing the performances of the proposed model, this research considered a few cancer patterns namely bladder, breast, colon, endometrial, kidney, leukemia, lung, melanoma, mom-hodgkin, pancreatic, prostate and thyroid. From our experimental results, it was noticed that the proposed model outperforms the identified three classifiers in terms of memory utilization, execution time (processing time), sensitivity, specificity, classification accuracy and F-score.