Analysing the Best Algorithm and Hyperparameter for Brain Tumour Detection

Early detection is necessary to achieve better patient outcomes because tumours are the second most common cause of cancer. Brain tumours can result in the formation of aberrant brain cells, some of which may evolve into cancer; hence, the medical field needs quick, automated, trustworthy, and efficient methods to diagnose them. An MRI scan is a typical way to find brain tumours. The accuracy and efficacy of identifying brain tumours using MRI data have recently increased because of advances in deep learning algorithms and machine learning. The proposed work intends to further determine the best algorithm and hyperparameter that enhance the accuracy of brain tumour identification from MRI images by applying logistic regression (LR), support vector machines (SVMs), K-Nearest Neighbours (KNNs), decision trees (DTs), and random forests (RFs) which are examples of machine learning approaches. In this research, we employed the entropy criterion, Z-score normalization, and hyperparameter tuning for better accuracy and evaluated the performance of each algorithm based on accuracy, accuracy, recall, and F1-score to produce more precise and effective results. In the case of accuracy, random forest came in first with a 99.067% accuracy rating, then came decision tree and K-Nearest Neighbour. The suggested methodology may help medical practitioners treat patients with brain tumours promptly and accurately, ultimately leading to better patient outcomes.