一种基于关联规则与支持向量机的基因表达数据分类模型

摘要

The discovery of association rules correlated to specific diseases through studying the gene expression data is of great importance to the computer aided diseases diagnoses.Here we propose an association rule-based gene expression data classification model (ASSO-SVM) aiming at the deficiency of interpretability in existing classification results.In this model,association rules are used as a kind of feature selection approach to extract the nonlinear associations among the genes.The priori knowledge acquired by these nonlinear associations benefits the improvement on interpretability of the classification results.Besides,in light of the features of high dimensionality and small sample of the gene expression data,the method uses support vector machine as the classifier for gene expression data to achieve higher classification accuracy.The ASSO-SVM combines the advantages of the gene expressive association rules and the SVM classification.Contrastive experiment on practical gene expression datasets in comparison with existing classification model verifies the effectiveness of the method.%通过研究基因表达数据发现与特定疾病相关的关联规则，对疾病辅助诊断有重要的意义。针对现有分类结果可解释性的不足，提出一种基于关联规则的基因表达数据分类模型ASSO-SVM（ASSOciation rule based Support Vector Machine）。在该模型中，关联规则作为一种特征选择方法，用于提取基因之间的非线性关联。通过这些非线性关联所获取的先验知识有利于提高分类结果的可解释性。另外，针对基因表达数据高维、小样本的特性，该方法采用支持向量机作为对基因表达数据的分类器，获得较高的分类精度。ASSO-SVM结合了基因表达关联规则以及支持向量机分类的优点。在实际基因表达数据集上与现有分类模型的对比实验验证了该方法的有效性。