CAJ | 학술논문

　　针对靶基因样本数据不平衡导致阳性样本预测准确率较低的问题，提出基于SVM的靶基因预测算法，即偏置判别SVM。算法选取高质量的数据集和最优特征集作为输入，在经验特征空间中以偏置判别分析准则为核优化目标函数，使用核保角变换的方法逐步优化核矩阵，用最优核矩阵构造偏置判别SVM，以解决靶基因数据不平衡对预测造成的影响。对比实验分析表明，提出的偏置判别 SVM 算法具有更高的特异度、敏感度和预测精度。同时，偏置判别SVM具有更强的泛化能力，鲁棒性更好。
　　침대파기인양본수거불평형도치양성양본예측준학솔교저적문제，제출기우SVM적파기인예측산법，즉편치판별SVM。산법선취고질량적수거집화최우특정집작위수입，재경험특정공간중이편치판별분석준칙위핵우화목표함수，사용핵보각변환적방법축보우화핵구진，용최우핵구진구조편치판별SVM，이해결파기인수거불평형대예측조성적영향。대비실험분석표명，제출적편치판별 SVM 산법구유경고적특이도、민감도화예측정도。동시，편치판별SVM구유경강적범화능력，로봉성경호。
For the data imbalance problem of miRNA target, a target prediction algorithm, Biased Discriminant Support Vector Machine, is proposed to solve the lower prediction accuracy of positive samples. The high-quality data sets and the optimal feature set are selected as the input data. Biased discriminate analysis criteria is selected as the kernel optimize objective function in the empirical feature space, and the conformal transformation of a kernel is adopted to gradually optimize the kernel matrix. Then, the SVM classifier with the optimal kernel matrix is constructed to solve the problem for the prediction causing by imbalance data. Through comparison with the analysis of the experimental results, the biased discriminant support vector machine method shows higher specificity, sensitivity and prediction accuracy, which proves that it has stronger generalization ability and better robustness.