基于数据挖掘方法的皮肤病诊断建模

Building Diagnostic Model for Dermatology Disease by Using Data Mining Methods

临床特征和组织病理学特征在皮肤科疾病分类问题中起着非常关键的作用。本文的主要目的是使用这些特征建立皮肤科疾病分类模型。本文首先使用最大相关最小冗余方法选择相关特征,然后使用支持向量机构造分类模型。文中所使用的数据集包含358皮肤病患者病例样本和35属性。通过实验比较使用特征选择方法前后的模型分类精度可知,使用特征选择方法能明显提升模型分类性能。同时,本文通过实验研究对不同年龄段的皮肤科疾病患者分类的差异性,并选择出了对于35岁以下皮肤科疾病患者分类重要的13个特征以及35岁以上患者分类重要的9个特征。

The clinical features and histopathological features play a key role in the performance of the dermatology classification problem. The primary aim in this work is building a dermatology classification model by using these features. In this paper, a well-known feature selection method, maximum relevance minimum redundancy （mRMR）, is used to select relevant features. Then, support vector machine （SVM）, a popular classifier is applied in the classification model. The data set used in this work contains 358 cases and 35 attributions including 34 features and 1 class. The experiment results demonstrate that the accuracy of classification model is improved obviously through feature selection. At the same time, 13 and 9 optimal features are selected from the data with patients aged under 35 and over 35 respectively. These features are regarded as the most significant features to classify the skin diseases for patients at different stages of age.