摘要： 卵巢癌蛋白质谱数据属于高通量数据，含有大量冗余信息，且许多重要信息都包含在高阶信息中，而独立成分分析可以从高阶信息中提取有用特征。将传统的独立成分分析融入卵巢癌蛋白质谱数据的特征提取中，并利用类信息机制监督独立成分分析过程。仿真实验结果表明，独立成分分析和监督式独立成分分析在卵巢癌蛋白质谱数据集研究过程中取得了良好的效果，识别率可达98%。

关键词: 独立成分分析, 特征提取, 散布度, 质谱数据

Abstract: Ovarian cancer protein mass spectrometry data is a high-throughput data and has large numbers of redundant information. The important information is contained in the high-order information. Independent Component Analysis(ICA) can extract useful features from high-order information. This paper makes ICA into the process of feature extraction for ovarian cancer protein mass spectrometry data, and uses class information mechanism to supervise the process of ICA. Simulation results show both ICA and Supervised Independent Component Analysis(SICA) have gained better effect in research of ovarian cancer protein mass spectrometry data, and the recognition rate reaches 98%.

Key words: Independent Component Analysis(ICA), feature extraction, scatter degree, mass spectrometry data

中图分类号: 

• TP18