Brain Tumor Classification from Gene Expression Dataset Through Supervised Machine Learning Algorithms

Brain tumors are amongst the most prevalent diseases affecting the brain, can cause severe pain, and lead to various illnesses if not adequately treated. A notable factor contributing to the rising number of cancer cases worldwide is the lack of awareness and early intervention in tumor treatment. Early-stage diagnosis becomes crucial due to nonspecific symptoms, resulting in an improved prognosis. The main motive of this work is to categorize brain tumors using the gene expression dataset. The Gene expression dataset used in this study was a Gene Expression dataset which consists of 7130 features that include the information of 40 patients. Since the dataset contains many attributes, Principal Component Analysis (PCA) and Pearson correlation are used to choose features to minimize dimensionality for modeling. Machine learning approaches are utilized to categorize various types of brain tumor. The categorization of brain tumors involves the use of machine learning models such as Decision Tree Classifier (DT), Support Vector Machine Classifier (SVM), Logistic Regression (LR), K-Nearest Neighbor Classifier (KNN), and Random Forest Classifier (RF). The concert of the machine learning approaches was assessed via classification performance metrics including accuracy, recall, precision, confusion matrix, and f1-score. Out of 40 patients, 10 patients were identified to have Medulloblastoma, 10 were found to have Glioma, 10 were found to have RhabdoidTu, 6 were found to have Primitive Neuro-Ectodermal Tumor (PNET) and 4 were found to be NORMAL. The experimental investigation concluded that the Logistic Regression approach outpaced other approaches such as RF, SVM, DT, and KNN.