多标签分类算法，通过对标签降维（SVD），然后利用线性回归建立特征和低维标签之间的关系，求出特征的系数，然后反过来进行预测(Multi label classification algorithm, through the tag dimension reduction (SVD), and then use linear regression to establish the relationship between features and low dimensional tags, to find the coefficients of the feature, and then in turn to predict)