CAJ | 학술논문

针对传统平行坐标只关注原始数据特征和分布情况的缺点，在平行坐标的基础上，提出了一种多维数据的基于相关规则余弦值的分类方法。该方法通过计算多维数据属性的支持度和余弦值，然后定义贡献度，根据构建的贡献度平行坐标对该数据集进行可视化数据分析以及对未知数据进行分类，从而获取数据中潜藏的有用信息。将该方法用于分析一个Pima印第安人糖尿病数据集，结果表明：利用贡献度平行坐标图表可以清楚地看出每个属性对最终分类结果的影响程度，得到比较满意的分类效果。
침대전통평행좌표지관주원시수거특정화분포정황적결점，재평행좌표적기출상，제출료일충다유수거적기우상관규칙여현치적분류방법。해방법통과계산다유수거속성적지지도화여현치，연후정의공헌도，근거구건적공헌도평행좌표대해수거집진행가시화수거분석이급대미지수거진행분류，종이획취수거중잠장적유용신식。장해방법용우분석일개Pima인제안인당뇨병수거집，결과표명：이용공헌도평행좌표도표가이청초지간출매개속성대최종분류결과적영향정도，득도비교만의적분류효과。
Aiming at the disadvantages that the traditional parallel coordinates only concerned with the characteristics and distribution of the raw data, on the basis of parallel coordinates, a classification method is proposed based on the relevant rules-cosine to deal with multidimensional data. The method calculates the support and cosine of multidimensional data attributes, and then defines contributions. According to build of the parallel coordinates of contributions, the data set can be visually analyzed and the unknown samples can be classified in order to receive the useful underlying information from the data set. This method is applied to a Pima Indian diabetes data set. The results show that by using the contributions parallel coordinates, the influence degree of each attribute on the last classification result is clear and the classification is satisfying.