Pembangunan ontologi kanser payudara bagi pemilihan data dalam meramal risiko

Abstract

Breast cancer is a deadly disease caused by the uncontrolled growth of cells that starts in the breast. Therefore, the accurate risk prediction is crucial in assisting the selection for the suitable prevention treatment, depending on the level of the risk. However, the abundance of biomedical data from various sources creates difficulty in data organizing. In addition, the big challenge in predicting the risk of breast cancer is the different attributes of the datasets which make it inscrutable for someone who are not from the domain background. Ontology is a new method introduced to improve the knowledge discovery in complex database. Ontology approach was applied in this study to resolve this problem by providing clearer understanding of the data. In this study, ontology was also used to select important features for data analysis. Classification technique of Sequential Minimal Optimization (SMO) was also applied in this study. SMO is a fast learning algorithm of Support Vector Machine (SVM) and able to provide high accuracy results. However, the analysis of breast cancer risk shows that data analysis without ontology has slightly higher accuracy compared to data analysis with ontology, where, the first dataset is 94.7% compared to 92.1% and the accuracy for the second dataset is 96.7% compared to 96.6%. These results were different from expectation, which the application of ontology was supposed to be able to provide higher accuracy results. This is caused by the limitation of data available for this study. Therefore, the study on breast cancer risk prediction by using ontology can be improved in the future by using broader cancer data and consistent cancer data type.