Classification Analysis of Some Cancer Types Using Machine Learning

Cancer is a disease caused by changes in deoxyribonucleic acid, which attacks cells in the body, causing them to grow uncontrollably and spread to other parts of the body. Cancer can be deadly. The fact that it can develop anywhere in the body gives rise to many types of cancer. Because a good diagnosis increases the probability of administering a good treatment to save life. Therefore, to reduce the mortality rate from cancer, several diagnostic methods have been developed as the appropriate treatment option is highly dependent on the type of cancer. In this work, we address the issue of classification of some cancer types by using supervised learning methods to classify prostate cancer, lymphoma, leukaemia and small round blue cell tumour. To be more specific, we used five models: support vector machine, decision tree, random forest, K-nearest neighbours (KNN) and artificial neural network. Each cancer dataset was trained using each of the machine learning methods on the Google Colab graphics processing unit (GPU). The test samples were classified for each cancer type, and the performances of the five models were compared in terms of their percentages according to some metrics. To reduce the dimension of the data, we have incorporated a new approach that involves performing principal component analysis on our dataset. This new approach led to the discovery that the KNN method was the best according to our dataset, with 90% accuracy for the prostate and 100% for the others.