Classification of cancer with microarray data is a challenging task because of the high dimensionality and imbal-anced characteristics of the dataset. Selection of few contributing relevant feature genes and classification of biomedical data with an optimally tuned classifier enhances the accuracy of disease detection. Generally, choice of parameters of the classifier is done by trial method and a value is kept fixed for the classification. But in this study during the search process, the best parameters of the classifier are adapted. Five well established and robust algorithms in the multi-objective domain are chosen for the feature selection and for the tuning of parameters of the support vector machine (SVM). We have performed experiments on 12 microarray datasets (both binary class and multi-class). Not only the effect of classifier tuning is studied but also a thorough comparison is performed amongst different multi-objective algorithms. The comparative study can be helpful to make an informed choice for selection of an appropriate algorithm in the field of multi-objective feature selection. The experiments show that MODE provides promising results for highly imbalance data having higher number of classes. For binary and multiclass data with imbalance ratio within the range of 2-5, NSGA-II is performing better. For the dataset having low imbalance ratio, the performance of the multi-objective algorithms are quite competitive. © 2021 IEEE.