Comparison of Colorectal Cancer Classification between K-Nearest Neighbors (K-NN) and Neural Network

Abstract

Machine learning is one of the technologies used in medicine. Machine learning can help detect various kinds of problems in the medical field and enables a process to be faster and more efficient. Cancer is one of the most dangerous diseases in the world. Machine learning is widely used in bioinformatics and particularly in cancer diagnosis. One of the most popular methods is K-nearest neighbors (K-NN) and Neural Network. There are supervised learning methods. Using K-NN, the quality of the results depends largely on the distance and the value of the parameter “k” which represents the number of the nearest neighbors. This research is explains the classification of colorectal cancer by using K-NN with different k values and Neural Network Classification. Our work will be performed on the Colorectal Cancer dataset obtained by the Al-Islam Hospital, Bandung, Indonesia and it consists of benign cases 163 and malignant cases 47 samples. Thus, the final result indicates better performance for K-nearest neighbors’ accuracy is 0.786 in K-parameter equal to 7, 9, 11 has the same accuracy with 60% data training and Neural Network reached 0.904 with 90% of data training.