Development of novel methodology for gene identification-based classification of leukaemia disorder

Purpose Many classifier approaches and algorithms are developed in recent days to handle large dimensionality problems in biomedical data mining and machine learning. The research focus to identify the optimum search methodology to predict the gene samples for leukaemia. Methods Different search classifier such T -test, Principal Component Analysis (PCA) and Genetic Algorithm (GA) are considered in this study and the search results of each classifiers are analysed. The classifiers are used to filter the top important and sort the mutually exclusive illness samples. The classifiers T -test and PCA are blended with Linear Discriminant Analysis (LDA), Self-Organizing Map (SOM) and Random Optimized Search (ROS) to predict the performance of the coupled classifiers. Results The confusion matrix is employed to calculate accuracy and compare the considered classifiers’ performance and accuracy. GA classifiers show a better performance than the other classifier-based feature selection algorithms with substantially unique gene characteristics. The mean, best and average generations of GA are considered to determine the accuracy of the generations. Conclusion The ROS-based LDA classifier improves the classification results and GA enhances the gene retrieval. The performance analyses of the different generations of GA are examined using the confusion matrix and the most optimal classifier is identified as GA-Avg-120G.