Intelligent Computing Approach in Gene Evaluation for Carcinogenic Disease Detection

This chapter addresses the challenge of identifying a small subset of crucial genes from microarray data, given the limited number of effective samples compared to the vast number of genes. The work suggests a novel method combining adaptive k-nearest neighbor-based gene selection with Particle Swarm Optimization (PSO) to obtain accurate identification of cancer subtypes. The aim is to identify a small yet informative set of genes that can be properly classified. In order to effectively explore the right neighborhood and accurate classification, the suggested method incorporates a heuristics for calculating the optimal value of k. Three benchmark microarray data sets namely SRBCT, ALL_AML, and MLL are used to test the suggested gene election method. The method’s effectiveness is demonstrated by the results, which include high classification accuracy on blind test samples, the number of informative genes identified, and computational efficacy. Furthermore, other classifiers, such as Support Vector Machine (SVM), are used to assess the usefulness and universal properties of the identified genes. The results confirm that the suggested methodology is resilient and useful in the field of carcinogenic disease classification.