Colorectal Cancer Risk Factors’ Assessment in Palestine: A framework for prediction tool رسالة ماجستير

Abstract

Healthcare is considered as one of the fields that produce tremendous amount of data, and this produced data will be useless if useful patterns are not extracted and used in a proper way. Generally, different types of cancer forms about 14% of mortality in Palestine, and Colorectal Cancer (CRC) prevalence especially scored 15% among men and 14.6% among women of all cancer types. On the other hand, CRC incidence depends relatively on behavioral risk factors that might increase CRC incidence and preventive factors that could decrease CRC incidence. Therefore, this research was carried out in order to spot the behavioral risk factors that affected Palestinian reported CRC cases and to make use of Machine Learning (ML) tools which might be used in CRC prediction, where the use of public CRC classification and prediction tool based on accurate ML tools will help individuals in tuning their behavioral CRC risk factors and enhance their engagement with their own health. In this research, two datasets were collected and analyzed two different CRC datasets, where one obtained from the National Cancer Registry of the Palestine Ministry of Health and the other dataset collected from the database of Al Quds University. The study found that behavioral the most important risk factors to consider are age, past medical history, diet behaviors, physical activity, and obesity. Consequently, different machine learning tools were applied to classify and predict CRC risk factors. In this research, machine learning tools were used to support medical decisions by combining intelligent computational systems with medical datasets. The task is to classify or recognize a different patterns in medical diagnoses to determine medical CRC risk factors. Decision Tree (DT), K-Nearest Neighbor (KNN), Support Victor Machine (SVM) and Artificial Neural Network (ANN) tools were applied. Therefore, this study determines a technique that can be used for diagnosing CRC risk factors. The Artificial Neural Network (ANNs) model outperforms other ML models with better accuracy in the two v collected datasets and it is the best. The ANNs model was applied to the two datasets. Then models’ performance was evaluated using different metrics such as the overall accuracy, recall, specificity, and AUC, and others. The obtained results show that the ANNs model outperformed all models applied in this thesis in the classification and prediction of CRC risk factors and their types. Finally, the study also found a crucial need to promote CRC preventive factors such as CRC screening and individuals’ awareness about CRC.