Performance Analysis of Leukemic Gene Expression Profiles using Classification

Evaluation and diagnosis of any cancer disease is facilitated with the classification of blood cells. Once blood disorders are identified it is easy to classify diseases related to blood. Leukemia is a cancer that affects the bone marrow, which is the producer of blood. Studies reveal that Leukemia can be classified using blood smear images or by taking gene expression profiles. Classification based on gene expressions is of much interest to researchers since it provides an objective, accurate and systematic diagnosis of several cancer types. The present study is an analysis of the classification of gene expression profiles from the classic Golub et al leukemia dataset which reported how gene expression data from DNA microarrays could be used to classify new cancer cases. The present study uses the K-Nearest Neighbors, Support Vector Machines, Naïve Bayes, Logistic Regression, Random Forest and Neural Network algorithms to classify the samples into either Acute Myeloid Leukemia (AML) or Acute Lymphocytic Leukemia (ALL). Performance evaluation of the various algorithms used is given towards the end of this paper. The results reveal that Logistic Regression provides accurate results and the study reiterates the fact that Neural Networks perform efficiently when the dataset is large.