Breast Cancer Gene Expression Analysis for the Classification of the Disease using Data Mining Algorithms

Breast cancer is the most incurable cancer and death rates growing with late-stage identification; early detection is strongly recommended to lower death rates. Cancer is defined as a disease resulting from altered gene expression, which is the process by which genetic material is utilized to produce useful molecules that regulate different aspects of cellular activity. Both substantial alterations in the regulation of genes and mutations are responsible for these alterations. Genes contain inherited genetic information that is passed down through the generations. Certain genetic defects that are also inherited by the next generation cause human diseases. In order to detect this information distortion, specific techniques are required. Several data mining methods are being researched to forecast breast cancer using information obtained from DNA microarrays regarding gene expression. In this study, cancer genes that can be used to support a patient's diagnosis are identified using data mining approaches such as Radial Basis Function Neural Network Algorithms (RBFNN) and Divide and Conquer Kernel Support Vector Machine Algorithm (DCKSVM). The results of the algorithms have been compared using a variety of confusion matrix metrics following the implementation phase. When compared to other algorithms already in use, the comparison results show that the (RBFNN) performs better, requires less memory. The significant variance in gene characteristics and a limited number of samples present an obstacle when using gene expression data. Therefore, a methodology to reduce the features is required. In this work, similar genes are combined using K-means clustering to pre-process the gene expression data.