Machine Learning for Colorectal Cancer Risk Prediction

Colorectal cancer is the third most prevalent cancer and the second most common cause of cancer deaths in the United States. Screening is one of the most powerful tools for colorectal cancer prevention. Current screening recommendations are only based on history of colorectal cancer and age. To facilitate a more effective screening of color ectal cancer, this paper explores the feasibility of machine learning algorithms for the colorectal cancer risk prediction. The longitudinal Pancreatic, Lung, Colorectal, Ovarian Cancer dataset from the National Cancer Institute was utilized for the training and testing of eight machine learning algorithms. The experiment results show that the gradient boosting model has the largest area under the Receiver Operating Characteristics curve 0.82, and the random forest model has the highest accuracy 0.75, highest recall 0.76 and highest F1 score 0.75. The two optimal models were also used to evaluate the importance of top risk factors, which are helpful for a more effective screening recommendation.