关键词: 服装部件, 裤脚, 个性定制, 感性工学, 映射模型, 个性化推荐

Abstract:

Objective In recent years, consumers' demand for clothing has turned to be more individualized, diversified and intelligent. The purpose of this paper is to help enterprises accurately grasp the emotional preference of consumers in the customization process, so as to match the personalized recommendation of product core components with users' emotional needs, thus achieving successful personalized customization. This paper takes the female tracksuit bottom as an example to establish a personalized recommendation model based on Kanseiengineering, and to help consumers customize personal schemes based on their needs.
Method Based on the principle of Kansei engineering, this paper firstly collected the bottom styles and design elements. Adjective words were selected to describe the style, and semantic difference method was used to obtain consumers' perceptual evaluation in seven dimensions and build a perceptual image space. Then, the author set up a mapping model between design elements and consumer perception through partial least square(PLS) method. Analytic hierarchy process (FAHP) was applied to quantify the perceptual needs of consumers, and a personalized recommendation model combining with mapping modelwas established.
Results Through a preliminary screening, literature review and expert consultation of 100 tracksuit bottom designs, the quantifiable factors wereset including looseness, closing method and slitting method. Morphological analysis was used to decompose the three elements twice to obtain 12 sub elements. The design elements and coding table are shown in Tab. 1. After preliminary screening, questionnaire survey and expert screening of 120 perceptual adjectives werecollected, 7 pairs of adjectives were finally obtained which were used to establish the perceptual image space of female tracksuit bottoms. The vocabulary and its definition angle are shown in Tab. 2. According to the principle of Kanseiengineering, a 7-level scale was designed by semantic difference method. 70 female college students with exercise habits were randomly invited for questionnaire survey, and 64 valid questionnaires were obtained. The average score of sample styles are shown in Tab. 3. Minitab software was used to conduct regression analysis on the average scores of style design elements and adjectives. The regression coefficient table is shown in Tab. 4. According to the regression coefficient, a mapping model between design elements and consumption perception was established. Through questionnaire, users wereasked to choose the perceptual image words of preference to describe individual needs. For example, user I's perceptual image acquisition and demand emphasis are shown in Tab. 5 and Tab. 6. The weight of perceptual image wascalculated by FAHP, thus obtaining user one's perceptual image weight expression. Based on weight, clothing set distance sortingwas adopted, and recommendations were made according to the sorting results. For the case of user I, the comprehensive evaluation distance P was sorted of each experimental sample, and four styles that meet the perceptual needs of user I were generate(Fig. 4). 15 consumers were invited at random again to make recommendations, and the recommendation results were obtained, and the consumers were asked to conduct emotional evaluation on the recommendation results. The similarity of the score matrix between the recommendation results and the perception evaluation was compared by calculating the cosine similarity of formula 4. The average similarity reached 0.902, which was relatively high. The average absolute value error (R_MAE) of formula 5 was used to evaluate the accuracy of recommendation results(Fig. 5). R_MAE was all less than 0.75. The recommendation algorithm was found able topredict and recommend accurately and has certain application value.
Conclusion Based on Kansei engineering, this paper proposed a personalized recommendation model for tracksuit bottom, and demonstrates the algorithm and process of the recommendation. Through testing, it shows that this model can effectively transform the emotional needs of consumers into design elements, so that the recommendation results can be matched with user needs, thus realizing personalized recommendation for tracksuit bottom based on the emotional needs of users, and improving the efficiency of personalized customization. At present, only the styles of female tracksuit bottoms have been evaluated and recommended. In the future, more comprehensive studies can be carried out based on fabric comfort and color. Besides, the tracksuit bottom is only one part of a garment, and the research object can be expanded to other parts.

Key words: apparel component, trousers, personalized customization, Kansei engineering, mapping model, personalized recommendation