基于独立成分分析(ICA)的分类研究及在蛋白分类中的应用

独立成分分析方法(ICA)一般用于特征提取、过程监控,很少用于分类。本文建立了一个ICA用于分类的工具包,并应用专家库下载的嗜热蛋白和常温蛋白作为实验数据,用来检验该分类工具的性能。本文的检验工作分为如下几个步骤:首先对训练数据进行特征提取;然后利用核函数确定嗜热蛋白的置信线;再计算测试数据的统计量。由测试数据判断是否超过置信线,不超出则认为是嗜热蛋白。该检验的结果是:对嗜热数据的漏检率为10%,误检率为30%,预测的正确率可达到80%。由此可说明,ICA可用作一种分类方法,本文所建立的分类工具也是有效的。

Study on Classification Based on ICA and Its Application in Protein Classification

Independent component analysis（ICA） are generally used for feature extraction,process monitoring,but rarely used for classification.A tool kit for classification was created,which could apply the thermophilic and mesophilic proteins download from the expert databases as the experimental data to test the performance of the classification tool.First of all,features of training data were extracted and Kernel function was used to determine the confidence line.Secondly,the amount of test data statistics was calculated.At last,determined whether it beyond the confidence line by the test data,if it beyond the confidence line,it could be considered as mesophilic proteins.The test results were： the missed rate for thermophilic proteins was 10%,the false detection rate was 30%,and the prediction accuracy was 80%.This can explain that ICA can be used as a classification method and the classification established by this tool is effective.