[學術演講] Classification Method - Finite Mixture Models for Early Cancer Detection
A Classification Method based on
Finite Mixture Models for Early Cancer Detection
We propose a new classification method under the framework of finite mixture models that can easily accommodate multiple biomarkers into a mixture and greatly increase the area under the ROC curve. With no assumption on the distributions of biomarkers, the proposed method can provide a more general framework for both discrete and continuous biomarkers and remarkably improve the accuracy of classification. An intensive simulation study suggests that the proposed method outperforms other existing methods with a higher correct classification rate. The data from a study of in-vitro assays based on the Fe/Fe3O4-based nanoplatforms for early cancer detection were used to illustrate the proposed methodology.