基于因子分析和聚类分析的情感维度提取∗

摘要

To extract the representative affective dimensions from a wide range of affective dimensions, an approach for selecting the representative image word pairs based on factor analysis ( FA) and cluster analysis ( CA) was proposed. Firstly, consumer’ s perceptions toward a small number of representative product sam-ples were obtained using semantic differential ( SD ) method. Secondly, the latent factors and the factor loading matrix of the initial affective dimensions were extracted using FA according the SD results. At last, ranking the distance of the initial affective dimensions to the center of gravity of each cluster group based on the results of CA. The image word pair with the shortest distance to the centroid was selected as the repre-sentative of the cluster. Application process and procedure of the proposed method were described by a case of CNC machine tools design. The case study results revealed that the representative affective dimensions were selected effectively and the overall structure was preserved using the proposed approach.%为从大量用户情感维度(感性词对)中提取少量具有代表性情感维度(感性词对),文章提出结合因子分析( factor analysis, FA)和聚类分析( cluster analysis, CA)的情感维度提取方法。首先通过语义差分( semantic differential, SD)实验获取用户对少量具有代表性样品的情感认知；然后使用FA对SD结果进行分析,获得初始情感维度的潜在因子及因子载荷矩阵；最后根据FA结果进行聚类分析,对初始情感维度到每个聚类几何中心的距离进行排序,距离最近的感性词对即为所提取的情感维度。以数控机床造型设计为研究案例,对该方法进行描述,结果表明,该方法能有效提取情感维度,并保留了初始情感维度的整体结构。