Early Detection and Prognosis of Brain Tumor from Micro Array Gene Data using Machine Learning Classifiers

Brain tumour is one of the most lethal illnesses affecting the brain. It typically begins as little benign cells that form on the interior of brain cells and progresses to tumour. Brain tumours are caused by the spread of somatic mutations that change gene expression. The DNA microarray method offers a standardized format for evaluating the levels of gene expression in thousands of different genes. Microarray technology's patterns of gene expression can be utilized to distinguish between cancers from various anatomical regions. A two level classification strategy is suggested because processing the microarray data requires processing very large amounts of data due to the problem of dimensionality. The genes or characteristics are dimensionally condensed in the first level using Particle Swarm Optimization and Cuckoo Search Algorithm technique. Five classifiers namely GMM, KNN, BLDC, FDC and SVM (RBF) are used to detect brain tumor. Classifiers performance are validated from the parameters such as F1 Score, ER and MCC. When comparing all classifiers, SVM (RBF) settled at the accuracy (92%). SVM (RBF) appears to be the most effective classifier for distinguishing brain tumor cells from normal cells based on gene expression patterns.