Random forest prognostic factor in colorectal cancer

In developing countries such as Indonesia, colorectal cancer cases in women are the third largest after breast cancer and cervical cancer, whereas, in men, cancer ranks second after lung cancer, followed by the third is prostate cancer. This study aims to determine the factors that affect the survival of colorectal cancer patients in the city of Makassar, Indonesia. The data used in this study including colon cancer patients diagnosed first in 2012 in 4 hospitals in Makassar City and observed survival until 2015. Predictor variables consisted of comorbidity, stage of cancer, age, treatment status, the location of cancer, sex, and history of metastasis of patients with colorectal cancer. The samples used in this study were as many as 38 cancer patients. In this study, we are using random forest which is an algorithm used in data classification through tree merging by training on sample data. Random Forest also an ensemble method consisting of several decision trees as classifiers. In a nutshell, the accuracy of this models can be justified by the value of classification by Area Under Curve (AUC) equal to 50%. Moreover, the most influential variable on the survival of colorectal cancer patient is a history of metastasis of colorectal cancer patient, cancer location and gender respectively.