Classifying Leukemia through DNA Expression Data Mining Techniques

The development of Microarray technology has provided vast amounts of data in various fields, including a cancer database which plays a crucial role in cancer prediction and recognition. This database includes expression levels of thousands of genes from several samples; however, the primary challenge lies in the mismatch between the dimensions of genes and samples. To tackle this issue, Artificial Intelligence algorithms, data mining methods, and computer science technologies are used to reduce the dimensionality of the matrix containing the expression levels of genes. After this reduction step, cancer classification or recognition takes place. The present study focuses on the binary classification of leukemia cancer with the aim of enabling scientists to classify leukemia samples into two classes (Acute Lymphocytic Leukemia and Acute Myelocytic Leukemia) in more efficient and accurate manner. To achieve this objective, various feature selection methods were applied and compared, leading to the adoption of the best dimensionality reduction algorithm, resulting in the highest classification accuracy using the minimum number of genes (features).