Colon Cancer Prediction Using Machine Learning in Jordanian Patients: Preliminary Results

Colon cancer remains a significant cause of cancer-related mortality globally, with early detection pivotal in enhancing patient survival rates. This study applies machine learning (ML) techniques to predict colon cancer using a dataset sourced from several hospitals across Jordan, featuring demographic data and complete blood count (CBC) test results. The methodology encompasses two primary phases: data preprocessing, which includes structuring the data, addressing missing values, and standardizing feature scales; and predictive analysis using various advanced ML models. These models were meticulously trained and validated within this region-specific context, employing a rigorous cross-validation strategy. The results, assessed through metrics such as precision, recall, and ROC curves, indicate that certain models, like the XGBoost classifier, achieved high performance, exemplified by an 80% precision and sensitivity rates, whereas others fell short in predictive effectiveness. Future efforts will focus on optimizing model parameters and further tuning to enhance the diagnostic capabilities of ML in the early detection of colon cancer.