CAJ | 학술논문

支持向量机是数据挖掘的新方法，由于其优秀的学习能力而得到了广泛的应用，但是，传统的支持向量机算法在处理大规模问题时存在训练时间过长和内存空间需求过大的问题，而应用多个支持向量机构成多分类器系统进行并行学习，是目前解决文本分类中大规模数据处理问题的一种有效方法。在分析传统并行算法的基础上，提出了一种改进的基于多支持向量机的并行学习算法，实验结果表明，采用该算法可使得分类效率得到较大的程度的提高，虽然，分类精度相对传统的方法略差，但是，在可接受的范围之内。
지지향량궤시수거알굴적신방법，유우기우수적학습능력이득도료엄범적응용，단시，전통적지지향량궤산법재처리대규모문제시존재훈련시간과장화내존공간수구과대적문제，이응용다개지지향량궤구성다분류기계통진행병행학습，시목전해결문본분류중대규모수거처리문제적일충유효방법。재분석전통병행산법적기출상，제출료일충개진적기우다지지향량궤적병행학습산법，실험결과표명，채용해산법가사득분류효솔득도교대적정도적제고，수연，분류정도상대전통적방법략차，단시，재가접수적범위지내。
Support Vector Machines (SVMs) is a new technique for data mining. It has wide applications in various fields and is a research hot pot of the machine learning field. However, SVMs needs longer training time and larger memory when it is applied to handling large-scale problems. It’s an effective way to solve large scale data processing in text classification with multiple clas-sifier systems composed by multiple support vector machine classifiers. Based on the analysis of traditional parallel algorithms, an improved algorithm based on multiple SVMs is proposed. The experimental results indicate that the new algorithm works well in precision and recall rate in the condition that the speeds of classification increases remarkably. Compared with traditional algo-rithms, the classified accuracy is lower but is within the range for acceptance.