Abstract: To improve the diagnosis accuracy in analog circuit, a SVM ensemble algorithm based on double disturbance mechanism of attribute and model parameter is proposed. Firstly, aattribute reduction algorithm based on chaos-ant optimization algorithm is presented under method of set-covering, to partition attribute space into many sub-spaces. And then in each subspace, the model parameter is disturbanced in the region of Low-Bias. Finally, the final Ensemble result is obtained by using two times of majority-voting method. The diagnosis instance indicates that the proposed algorithm performace better than many other algorithms, take multi-SVM, AB algorithm, Bagging and so on for example.