Journal of Textile Research ›› 2020, Vol. 41 ›› Issue (10): 122-131.doi: 10.13475/j.fzxb.20191200510

• Apparel Engineering • Previous Articles     Next Articles

Design and realization of a collocation recommendation system for women's clothing

GAN Meichen1, LI Min1,2,3()   

  1. 1. College of Fashion and Design, Donghua University, Shanghai 200051, China
    2. Key Laboratory of Clothing Design and Technology, Ministry of Education, Donghua University, Shanghai 200051, China
    3. Shanghai Style Fashion Design & Value Creation Collaborative Innvoation Center, Donghua University, Shanghai 200051, China
  • Received:2019-12-02 Revised:2020-07-15 Online:2020-10-15 Published:2020-10-27
  • Contact: LI Min E-mail:fidlimin@dhu.edu.cn

RichHTML

16

PDF (PC)

339

Abstract:

In order to meet the huge demand of consumers for clothing collocation recommendation and make up for the lack of existing services, the women's fashion brand L was taken as a case study. With reference to the basic procedures of Kansei engineering, based on the literature research and design elements and style features of the brand, a Kansei image evaluation scale of fashion style and a classification table of design elements were established. Through the questionnaire survey and data analysis, the influence direction and degree of each design element on the Kansei image of fashion style were found and quantitative models of fashion style were constructed. On the basis of theoretical research and market research, clothing collocation rules were established. Combined with the quantitative models of fashion style, a women's clothing collocation recommendation system was developed, and the recommendation results were verified. The verification results show that the precision rate, the recall rate and the comprehensive evaluation index were all within a reasonable range, indicating that the system could effectively recommend the users' favorite products. Interviews with experimental users show that the system can basically meet consumers' demand for clothing collocation recommendation.

Key words: clothing collocation, Kansei engineering, quantitative fashion style, recommendation system, e-commerce

CLC Number: 

  • F724.6

Tab.1

Kansei image adjective pair"

序号 形容词对 描述角度
1 职业的—休闲的 穿着场合
2 女性化的—男性化的 性别
3 成熟的—年轻的 年龄
4 简约的—繁复的 构成
5 潮流的—大众化的 时尚度
6 轻快的—沉稳的 气质

Fig.1

Meaning of each evaluation value in 7th order evaluation scale"

Fig.2

Classification example of clothing silhouette. (a) H type; (b) A type; (c)X type"

Tab.2

Classification of design elements"

款式 色彩 面料
廓形 领型 袖型 腰线 衣长/裤
长/裙长		 门襟 纹样
图案		 细节
造型
H型、
A型、
X型		 无领、立领、翻领、驳领、连帽领 无袖、圆袖、插肩袖、落肩袖 无腰、中
高腰		 短、中、长 无门襟、单排扣、双排扣、拉链、暗门襟 无图案、条纹、格子、花卉、字母、几何图形 无细节造型、系带、荷叶边、抽褶、百褶、镂空、褶裥、流苏、蝴蝶结 红色系、黄色系、绿色系、蓝色系、白色、灰色系、黑色、棕色系 透薄型、厚重型、光泽型、柔软型、硬挺型

Fig.3

Examples of investigation sample images. (a)Coat; (b)Suit; (c)Shirt"

Fig.4

Fashion style's Kansei image evaluation system"

Tab.3

Summary of fitting process"

模型
编号		 复相关
系数R		 判定系
数R2		 调整后
R2		 德宾-沃森
统计量
1* 0.841 0.811 0.782 1.738
2* 0.826 0.792 0.763 1.784
3* 0.874 0.829 0.797 1.802
4* 0.898 0.758 0.727 1.658
5* 0.831 0.782 0.748 1.726
6* 0.819 0.783 0.754 1.764

Tab.4

Partial regression coefficients summary"

类型 1* 2* 3* 4* 5* 6*
常量 -0.266 -0.717 0.494 -2.483 0.075 -0.862
廓形 A型 0.043 -0.255 0.054 0.263 -0.417 -0.126
X型 -0.667 -0.371 -0.414 0.035 0.194 0.367
领型 无领 -0.415 -0.026 0.043 -0.110 -0.110 -0.145
立领 -0.345 0.133 0.364 0.084 -0.325 -0.192
翻领 -0.514 -0.022 0.207 0.187 -0.330 -0.442
驳领 -1.012 0.086 -0.267 0.159 -0.402 -0.003
连帽领 0.658 0.636 1.158 0.597 -0.644 -0.307
袖型 无袖 1.278 0.494 0.126 -0.028 -0.629 -0.189
圆袖 1.033 0.731 0.165 -0.053 -0.399 0.011
插肩袖 1.234 0.362 0.147 -0.030 -0.391 -0.088
落肩袖 1.745 0.970 0.654 -0.095 -0.522 -0.267
腰线 无腰 -0.453 0.062 -0.284 -0.121 0.444 0.327
衣长 短 0.127 -0.367 0.309 0.068 -0.104 -0.150
中 -0.040 -0.269 0.039 0.074 -0.103 0.003
门襟 无门襟 -0.330 -0.989 -0.198 0.382 0.070 -0.227
单排扣 -1.310 -0.993 -0.743 0.156 0.258 0.259
双排扣 -0.915 -0.883 -0.994 0.276 0.372 0.458
拉链 0.016 -0.332 0.221 0.423 -0.264 -0.297
暗门襟 -1.058 -0.886 -0.858 0.379 0.106 0.290
纹样
图案		 无图案 -0.198 -0.157 -0.181 -0.140 0.229 0.204
条纹 0.160 -0.106 0.241 0.399 -0.442 -0.200
格子 1.449 -0.373 1.097 0.652 -0.839 -0.749
花卉 0.387 -0.251 -0.153 0.697 -0.056 -0.011
字母 0.943 0.453 0.836 0.224 -0.076 -0.139
几何图形 0.405 0.068 0.124 0.436 -0.242 -0.107
细节
造型		 无细节 0.023 0.344 -0.215 -0.436 0.400 0.199
系带 0.108 0.054 0.048 0.217 -0.294 -0.036
荷叶边 -0.085 -0.204 -0.130 0.300 -0.127 -0.050
抽褶 0.098 0.218 -0.193 0.097 0.292 0.155
百褶 0.219 -0.456 -0.142 0.778 -0.393 -0.324
镂空 -0.145 -0.180 -0.320 -0.072 0.026 0.203
褶裥 -0.087 0.455 -0.200 -0.071 0.053 0.110
蝴蝶结 -0.043 -0.387 -0.130 0.069 -0.086 -0.330
流苏 0.797 0.336 0.307 0.137 -0.261 -0.151
色彩 红色系 0.338 -0.601 -0.043 0.415 -0.270 0.188
黄色系 0.668 -0.072 0.188 0.076 -0.024 -0.307
绿色系 0.286 0.084 -0.313 0.591 -0.126 0.536
蓝色系 0.441 0.008 -0.076 0.511 -0.150 0.584
灰色系 0.284 0.262 -0.422 0.413 0.070 0.667
黑色 -0.009 -0.018 -0.454 0.478 0.056 1.180
棕色系 0.169 -0.065 -0.315 0.436 -0.242 0.777
面料 透薄型 -0.067 -0.200 -0.133 0.223 0.034 -0.118
厚重型 0.020 0.023 0.057 0.373 -0.290 0.029
光泽型 0.286 0.111 -0.027 0.171 -0.231 0.013
柔软型 -0.176 -0.136 -0.481 0.145 0.130 0.138
硬挺型 -0.029 0.018 -0.058 0.292 0.211 -0.044

Tab.5

Collocation rules of women's clothing category"

品类 可搭配品类
上装 大衣 衬衫、套头衫、裤子、半裙、连衣裙/连体裤
西装 衬衫、套头衫、裤子、半裙、连衣裙/连体裤
外套 衬衫、套头衫、裤子、半裙、连衣裙/连体裤
衬衫 大衣、西装、外套、裤子、半裙
套头衫 大衣、西装、外套、裤子、半裙
下装 裤子 大衣、西装、外套、衬衫、套头衫
半裙 大衣、西装、外套、衬衫、套头衫
连身装(连衣裙/裤) 大衣、西装、外套

Tab.6

Collocation rules of women's clothing color"

色彩 可搭配色彩
红色系 红、棕、灰色系,白、黑色
黄色系 黄、绿、棕、灰色系,白、黑色
绿色系 黄、绿、蓝、棕、灰色系,白、黑色
蓝色系 绿、蓝、棕、灰色系,白、黑色
棕色系 红、黄、绿、蓝、棕、灰色系,白、黑色
灰色系 红、黄、绿、蓝、棕、灰色系,白、黑色
白色 红、黄、绿、蓝、棕、灰色系,白、黑色
黑色 红、黄、绿、蓝、棕、灰色系,白、黑色

Tab.7

Women's clothing collocation recommendation process"

步骤 详细内容
①输入 用户A所选服装p、其余服装商品列表f={f1,f2,…}、所有服装的设计要素属性信息。
②筛选 步骤1:根据用户A所选服装p的品类与色彩属性,通过品类与色彩的搭配关联规则对其余商品fi进行初步筛选,排除不可搭配商品,得到待推荐商品列表J={J1,J2,…};
步骤2:依据商品p与Ji的设计要素属性信息,基于服装风格量化模型计算商品P与Ji的风格预测评价值,分别记为y^1,y^2,…,y^6与y^i1,y^i2,…,y^i6;
步骤3:根据欧几里得距离公式计算商品p与Ji的风格差异程度d(y^,y^i);
步骤4:根据d(y^,y^i)的大小对待推荐商品进行升序排列,以Top-N方法将各品类风格相似度(即风格匹配程度)最大的N件商品推荐给用户。
③输出 各品类的N件可搭配单品推荐将显示在服装p购买页面的“精选搭配”板块中。

Fig.5

System initial page"

Fig.6

Product purchase page"

Tab.8

Average of recommended results"

推荐列表长度N 准确率P/% 召回率R/% 综合评价F/%
3 78.9 21.2 33.4
6 76.1 40.9 53.1
9 75.6 60.7 67.2
12 71.9 77.2 74.4
15 64.9 87.0 74.2
[1] 王玮, 卜览, 廖念玲, 等. 重新定义新零售时代的客户体验|麦肯锡2017中国数字消费者研究[R/OL]. 2017- 06- 23[2018-12-31]. http://www.mckinsey.com.cn.
WANG Wei, BU Lan, LIAO Nianling, et al. Redefine customer experience in the new retail era|McKinsey 2017 China iConsumer survey [R/OL]. 2017- 06- 23[2018-12-31]. http://www.mckinsey.com.cn.
[2] 李玉莲. 基于专家系统的个性化服装搭配推荐系统的研究[D]. 北京:北京服装学院, 2016: 68-73.
LI Yulian. Expert system personalized recommendation system based outfit[D]. Beijing:Beijing Institute of Fashion Technology, 2016: 68-73.
[3] MAO Qingqing, DONG Aihua, MIAO Qingying, et al. Intelligent costume recommendation system based on expert system[J]. Journal of Shanghai Jiaotong University (Science Edition), 2018,23(2):227-234.
[4] MA Yihui, JIA Jia, ZHOU Suping, et al. Towards better understanding the clothing fashion styles: a multimodal deep learning approach[C]// Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence. San Francisco: AAAI, 2017: 38-44.
[5] 李砚祖. 设计新理念:感性工学[J]. 新美术, 2003(4):20-25.
LI Yanzu. New design concept:Kansei engi-neering[J]. New Arts, 2003(4):20-25.
[6] 钟厦, 宋晓晨, 孙亚云. 感性工学中量化的基本程序介绍[C]// 湖北省机械工程学会,中国机械工程学会湖北工业设计研究所,武汉科技大学,等. 2005年工业设计国际会议论文集. 杭州:中国机械工程学会工业设计分会, 2005: 728-731.
ZHONG Xia, SONG Xiaochen, SUN Yayun. Introduction of quantization's basic procedures in Kansei enginee-ring[C]// Mechanical Engineering Society of Hubei Province,Hubei Industrial Design Institute of Chinese Mechanical Engineering Society,Wuhan University of Science and Technology,et al. Proceedings of the 2005 International Conference on Industrial Design.Hangzhou:Industrial Design Chapter of China Mechanical Engineering Society, 2005: 728-731.
[7] 陈文勤. 基于感性工学理论的服装风格量化与建模研究[D]. 上海:东华大学, 2012:33-35,40-42.
CHEN Wenqin. The research of quantitating and modeling of fashion style based on Kansei engi-neering[D]. Shanghai:Donghua University, 2012:33-35,40-42.
[8] 黄俊敏, 李响, 孙莉, 等. 基于感性工学的男式衬衫风格量化研究[J]. 辽宁工程技术大学学报(社会科学版), 2016,18(6):953-960.
HUANG Junmin, LI Xiang, SUN Li, et al. Quantitative research on men's shirt style based on Kansei engineering[J]. Journal of Liaoning Technical University(Social Science Edition), 2016,18(6):953-960.
[9] 邵丹, 朱莉思. 基于眼动实验的服装品牌风格意象认知探析: E品牌上装风格案例研究[J]. 东华大学学报(自然科学版), 2013,39(2):240-246.
SHAO Dan, ZHU Lisi. Research on image cognitive of branded apparel based on eye-tracking technology: case study of E brand product style analysis[J]. Journal of Donghua University(Natural Science), 2013,39(2):240-246.
[10] 董向芳. 基于感性工学的女装设计研究与应用[D]. 杭州:浙江理工大学, 2013: 24-31.
DONG Xiangfang. The research and practice of dress design based on Kansei engineering[D]. Hangzhou: Zhejiang Sci-Tech University, 2013: 24-31.
[11] 彭荟. 论服装风格与设计要素之间的关联性[J]. 艺海, 2015(8):101-102.
PENG Hui. The relationship between fashion style and design elements[J]. Yi Hai, 2015(8):101-102.
[12] WANG Y, CHEN Y, CHEN Z G. The sensory research on the style of women's overcoats[J]. International Journal of Clothing Science & Technology, 2008,20(3):174-183.
[13] 李俊. 服装商品企划学[M]. 北京: 中国纺织出版社, 2005: 107-123.
LI Jun. Clothing commodity planning[M]. Beijing: China Textile & Apparel Press, 2005: 107-123.
[14] 张文彤, 董伟. SPSS统计分析高级教程[M]. 3版. 北京: 高等教育出版社, 2018: 102.
ZHANG Wentong, DONG Wei. Advanced course of SPSS statistical analysis[M]. 3rd ed. Beijing: Higher Education Press, 2018: 102.
[15] 卢纹岱. SPSS for Windows统计分析[M]. 3版. 北京: 电子工业出版社,2008:297- 312, 382-383.
LU Wendai. SPSS for windows statistical analysis [M]. 3rd ed. Beijing: Electronic Industry Press,2008:297- 312, 382-383.
[16] 罗娟, 吴奕苇. 服装搭配TPO原则与混搭风格之比较[J]. 广西轻工业, 2011,27(6):100-101, 144.
LUO Juan, WU Yiwei. Comparison of TPO principles and mixing styles in clothing collocation[J]. Guangxi Journal of Light Industry, 2011,27(6):100-101, 144.
[17] 朱宁, 陈寒佳. 服装色彩与搭配[M]. 合肥: 合肥工业大学出版社, 2015: 1-52.
ZHU Ning, CHEN Hanjia. Clothing color and matching[M]. Hefei: Hefei University of Technology Press, 2015: 1-52.
[18] 刘建国. 个性化推荐系统评价方法综述[J]. 复杂系统与复杂性科学, 2009,6(3):1-10.
LIU Jianguo. Overview of the evaluated algorithms for the personal recommendation systems[J]. Complex Systems and Complexity Science, 2009,6(3):1-10.
[1] . Application effect of blue and white patterns of Ming Dynasty on cheongsam [J]. JOURNAL OF TEXTILE RESEARCH, 2018, 39(01): 126-132.
[2] . Sensibility assessment of spring and summer shirt yarn-dyed fabrics [J]. JOURNAL OF TEXTILE RESEARCH, 2016, 37(08): 59-64.
[3] . Individual differences in perceptual evaluation for fashion design: taking female students as research subjects [J]. JOURNAL OF TEXTILE RESEARCH, 2014, 35(5): 137-0.
[4] WANG Ying;CHEN Yan. Relationship between perception and designing elements of woman′s overcoats [J]. JOURNAL OF TEXTILE RESEARCH, 2007, 28(11): 97-100.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
京ICP备14006648号
Copyright 2021 © Editorial office of Journal of Textile Research
Supported by Beijing Magtech Co.Ltd