中文摘要：

针对基于概率统计的ML—kNN算法只能对每个独立的标签进行分析，忽略了真实世界中标签间的相关性，提出了一种联系标签相关性的ML-kNN算法（S-ML—kNN）。该方法对训练集进行扩展，并按照标签间的二阶组合来构造新的标签，融合了标签之间的相关性。实验结果表明，S-ML-kNN算法优于ML-kNN算法。

英文摘要：

The only ML-kNN algorithm based on probability and statistics for each individual tag analysis, ignoring the corre- lation between the tags in the real world, this paper proposed a ML-kNN algorithm with the label correlation（ S-ML-kNN）, the method extended the training set and follow the label between the second combination to construct a new label, the integration of the correlation between labels. Experimental results show that, S-ML-kNN algorithm outperforms ML-kNN algorithm.