Cancer Prognosis Prediction Using SVM for Hybrid Type and Imbalanced Data Sets

Yanping Chen, Bingyu Su, Le Zou, Xiaoxuan Wu, Songhua Hu

Abstract

Cancer prognosis is one of the hot spots in the study of biological information. There have been many studies to cancer prognosis prediction using machine learning methods, which have achieved better results. Among them, the support vector machine (SVM) gets extensive attention as it is suitable to apply in small-size, high-dimensional data classification questions. However,SVM only performs well in the case where the class distribution is balanced and the input variables are numerical which are unlikely occurred in the medical domain. So in this study, we introduce a new prognosis prediction method based on SVM, which modify the standard SVM models to fit imbalanced class distribution and hybrid type of features. In details, firstly the similarity of features with nominal and numerical type is redefined in kernel function. Secondly synthetic minority oversampling technique (SMOTE) method is adopted to balance class distribution. Lastly the wrapper method SVM-RFE is introduced to select the useful features to improve the prediction performance. A series of experiments are designed and launched to validate the performance. The results have proved the effectiveness of the proposed methods.

Download


Paper Citation


in Harvard Style

Chen Y., Su B., Zou L., Wu X. and Hu S. (2018). Cancer Prognosis Prediction Using SVM for Hybrid Type and Imbalanced Data Sets .In 3rd International Conference on Electromechanical Control Technology and Transportation - Volume 1: ICECTT, ISBN 978-989-758-312-4, pages 13-18. DOI: 10.5220/0006964100130018


in Bibtex Style

@conference{icectt18,
author={Yanping Chen and Bingyu Su and Le Zou and Xiaoxuan Wu and Songhua Hu},
title={Cancer Prognosis Prediction Using SVM for Hybrid Type and Imbalanced Data Sets },
booktitle={3rd International Conference on Electromechanical Control Technology and Transportation - Volume 1: ICECTT,},
year={2018},
pages={13-18},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0006964100130018},
isbn={978-989-758-312-4},
}


in EndNote Style

TY - CONF

JO - 3rd International Conference on Electromechanical Control Technology and Transportation - Volume 1: ICECTT,
TI - Cancer Prognosis Prediction Using SVM for Hybrid Type and Imbalanced Data Sets
SN - 978-989-758-312-4
AU - Chen Y.
AU - Su B.
AU - Zou L.
AU - Wu X.
AU - Hu S.
PY - 2018
SP - 13
EP - 18
DO - 10.5220/0006964100130018