渤海大学学报(自然科学版)
渤海大學學報(自然科學版)
발해대학학보(자연과학판)
JOURNAL OF BOHAI UNIVERSITY(NATURAL SCIENCE EDITION)
2014年
1期
39-42,70
,共5页
超椭球%噪音点%协方差矩阵%马氏距离
超橢毬%譟音點%協方差矩陣%馬氏距離
초타구%조음점%협방차구진%마씨거리
hyper ellipsoidal%noises%covariance matrix .mahalanobis distance
提出一种基于马氏超椭球学习机的多类文本分类算法。对每一类训练样本,训练马氏超椭球学习机,使其包含该类尽可能多的样本,同时将噪音点排除在外。对于待分类样本,通过待分类样本的映射到每个超椭球球心的马氏距离确定其类别。实验结果表明,该算法提高了分类精度和分类速度。
提齣一種基于馬氏超橢毬學習機的多類文本分類算法。對每一類訓練樣本,訓練馬氏超橢毬學習機,使其包含該類儘可能多的樣本,同時將譟音點排除在外。對于待分類樣本,通過待分類樣本的映射到每箇超橢毬毬心的馬氏距離確定其類彆。實驗結果錶明,該算法提高瞭分類精度和分類速度。
제출일충기우마씨초타구학습궤적다류문본분류산법。대매일류훈련양본,훈련마씨초타구학습궤,사기포함해류진가능다적양본,동시장조음점배제재외。대우대분류양본,통과대분류양본적영사도매개초타구구심적마씨거리학정기유별。실험결과표명,해산법제고료분류정도화분류속도。
A new multiclass text classification algorithm based on mahalanobis hyper ellipsoidal learning ma -chine is proposed .To each class sample , training the hyper ellipsoidal learning machine , which include as much the class samples as possible and push the outlier samples away .For the sample to be classified , the distance from the sample mapping to the center of each mahalanobis hyper ellipsoidal are used to decide the sample classs.The results of the experiment show that the proposed algorithm has a higher classification accuracy and classification speed .