机器学习算法在药物专利分类中的应用研究

为实现药物专利的自动分类,本文结合药物专利的特点,研究了机器学习算法如何应用于药物专利分类。将2000余份药物专利按疗效分类,选取其中五类作为训练样本,对每一类提取特征文本,利用向量空间模型将非结构化的文本数字化,用支持向量机、Naive Bayes、RBFNetwork三种机器学习算法,分别测试专利样本的分类,使用5重交叉验证比较了三种算法的查准率(precision)和召回率(recall),结果表明支持向量机的分类效果最好。将机器学习算法应用于药物化学专利分类,有助于提高药物化学专利信息检索的效率。

Application of machine learning algorithms to pharmaceutical patent categorization

The applications of several machine learning methods to the categorization of pharmaceutical patents were presented in this paper. About 2000 pieces of pharmaceutical patents were categorized into five classes according to the curative effects and were selected as training instances. Features in text form were first extracted from each class and then were expressed in numerical vector form. Three machine learning algorithms, i. e. , SVM, Naive Bayes and RBFNetwork were evaluated using two most commonly used performance measures, precision and recall by a series of experiments. Results of five fold cross validation show that SVM outperforms than the other two algorithms. Methods proposed in this paper maybe helpful to the pharmaceutical patent categorization.