论文部分内容阅读
6类红斑鳞状皮肤病的诊断一直是皮肤科的难题。皮肤病数据是名词性定性数据,采用定量数据处理方法是不太适合的。本文提出了组套索罚多值回归分类器新方法用于名词性数据的特征选择和分类,并应用于红斑鳞状皮肤病诊断。首先将前33维名词性数据进行虚拟编码,将第34维年龄数据离散化后进行虚拟编码;将得到的虚拟编码数据按照类别分组和变量分组,并送入组套索罚多值回归分类器,通过10折交叉验证,分类正确率达到了98.88%±0.002 3%。与其他文献方法相比,本文方法简单,分类效果好且效率高,可解释性强,稳定性强。
Diagnosis of erythema scaly skin disease has always been a dermatological problem. Dermatology data is nominal qualitative data, the use of quantitative data processing methods are not suitable. In this paper, a new method of multi-value regression classifier for combinatorial punishment is proposed for feature selection and classification of nominal data and applied to the diagnosis of erythema and squamous skin diseases. Firstly, the first 33 dimensions of nominal data are virtually encoded, and the 34th-dimensional age data are discretized and then virtually encoded. The obtained virtual encoded data is grouped by category and variable, and sent to a set of multicollinear regression classifier , By 10 fold cross-validation, classification accuracy rate of 98.88% ± 0.002 3%. Compared with other literature methods, the method is simple, the classification effect is good and the efficiency is high, the interpretability is strong and the stability is strong.