Analysis of Different Feature Selection Techniques on Different Machine Learning Approaches to Classify Leukemia Subclasses

The correct identification of specific cancer subclasses is one of the significant challenges in the treatment of leukemia. Utilizing genomic profiles from microarrays is a useful tool for classifying different cancer subtypes. It is difficult since using vast amounts of gene expression data will demand a lot of processing power. The findings in our study focus on comparing several feature selection techniques in order to identify the genetic expressions that are extremely lethal. The feature selection procedure improves the algorithms’ capacity for prediction. The computational cost and data volume can be significantly reduced by choosing the appropriate feature. The correct identification of specific cancer subtypes is one of the major difficulties in treating leukemia. Utilizing genetic profiles from microarrays is a useful tool for classification. To classify Acute Myeloid Leukemia (AML) and Acute Lymphoblastic Leukemia (ALL) on features chosen in three different ways, namely the Filter method, Wrapper method, and Embedded method, six machine learning algorithms, and an Artificial Neural Network was constructed. One process achieved a success rate of 100% and the rest of the procedures procured about 95% accuracy on average.