The paper introduces author's job on realization of continuous attribute discretization based on languagefield theory that Prof. Yang put forward. It applies a new algorithm of seeking border values and its increm...The paper introduces author's job on realization of continuous attribute discretization based on languagefield theory that Prof. Yang put forward. It applies a new algorithm of seeking border values and its incremental onerather than seeking boundary ones that is a difficulty. The theory of the algorithm is self-contained, and its realiza-tion is simple. And the paper introduces simply four thoughts about defining language values and then discretizing fornon-numerical value that author already realized in KDD * .展开更多
连续属性的离散化是文本分类任务中数据预处理阶段的一项重要技术。针对机器学习领域中的诸多优秀算法只能处理离散属性的特点,提出一种基于词出现和信息增益相结合的多区间连续属性离散化方法(multi-interval discretization based on ...连续属性的离散化是文本分类任务中数据预处理阶段的一项重要技术。针对机器学习领域中的诸多优秀算法只能处理离散属性的特点,提出一种基于词出现和信息增益相结合的多区间连续属性离散化方法(multi-interval discretization based on term presence and information gain,MTPIG)。并将MTPIG算法应用到了分类算法AdaBoost.MH中,给出实验结果及分析。实验结果表明,使用MTPIG算法处理文本分类中的数据,其过程简单高效,预测精度高,可理解性强。采用该算法离散化数据集后,分类算法的分类性能有所提高。展开更多
文摘The paper introduces author's job on realization of continuous attribute discretization based on languagefield theory that Prof. Yang put forward. It applies a new algorithm of seeking border values and its incremental onerather than seeking boundary ones that is a difficulty. The theory of the algorithm is self-contained, and its realiza-tion is simple. And the paper introduces simply four thoughts about defining language values and then discretizing fornon-numerical value that author already realized in KDD * .
文摘连续属性的离散化是文本分类任务中数据预处理阶段的一项重要技术。针对机器学习领域中的诸多优秀算法只能处理离散属性的特点,提出一种基于词出现和信息增益相结合的多区间连续属性离散化方法(multi-interval discretization based on term presence and information gain,MTPIG)。并将MTPIG算法应用到了分类算法AdaBoost.MH中,给出实验结果及分析。实验结果表明,使用MTPIG算法处理文本分类中的数据,其过程简单高效,预测精度高,可理解性强。采用该算法离散化数据集后,分类算法的分类性能有所提高。
