Applied Soft Computing Journal 95 (2020) 106517 Contents lists available at ScienceDirect Applied Soft Computing Journal journal homepage: www.elsevier.com/locate/asoc Social image mining for fashion analysis and forecasting Seema Wazarkar ∗ , Bettahally N. Keshavamurthy Department of Computer Science and Engineering, National Institute of Technology Goa, Farmagudi, Ponda 403401, Goa, India article info Article history: Received 25 June 2019 Received in revised form 20 April 2020 Accepted 29 June 2020 Available online 6 July 2020 Keywords: Image clustering Classification Fashion analysis and forecasting Social images abstract Fashion industries need to be attentive towards the changing fashion and its upcoming market demands to grow their business, optimally. This paper describes research work involved in image mining for fashion analysis and forecasting using fashion-related images collected from the social network. A novel soft clustering technique is proposed for grouping the social fashion images. This technique is robust against uncertainty found in given images. The proposed clustering approach is compared with existing soft clustering approaches. It is found that the proposed approach performs well. Attributes of fashion items found in each cluster are analyzed through correlation, causal analysis, and fashion cycle visualization. Predictive models are applied to the clustered fashion items for style forecasting. A comparative study of predictive models is also done to find an optimal technique for various fashion items. As social visual perception is helpful for decision making, the proposed system is very useful in fashion industries to uplift their business. © 2020 Elsevier B.V. All rights reserved. 1. Introduction Fashion is followed by people from different cultures and places in their own way [1]. ‘‘If any chosen style or way of behaving is socially appropriate for a particular time period and situation, and a noticeable number of members from a social group starts adopting it, then that style of product or behav- ior is called Fashion’’ [2]. For example- baseball cap, straw hat, bowler hat, sun hat, cowboy hat, top hat, sombrero, turban, hijab, bonnet, etc. are the styles of headgear or headwear. Whereas, use of WhatsApp messenger, taking selfie pictures, etc. are the fashions in behavior. The word ‘‘Fashion’’ is used interchangeably with ‘‘Trend’’ or ‘‘Style’’ by many people. The trend represents a general direction or movement [3]. Style is a characteristic way of presentation that symbolizes several similar objects from the same category [2,4]. Everyone wants to be the first consumer for a unique and new fashion. Wearing things according to the latest style makes us representable and creates a good impression, therefore fashion is important in our day to day life. It changes frequently as people soon get bored with the prevailing fashion [5]. Fashion is time-dependent. It changes continuously due to the influence of world events, social and subcultural activities, economic con- ditions, technical advances, popular personalities in the society and fashion leaders like designers and celebrity icons. Therefore, fashion analysis and forecasting is complex and difficult but cru- cial task for the fashion industry to grow. It is a starting point ∗ Corresponding author. E-mail address: wazarkarseema@nitgoa.ac.in (S. Wazarkar). in the product development process. The fashion analysis and forecasting process includes three key elements i.e. environment, market and product as shown in Fig. 1. Every element follows five research process stages such as awareness and observation, searching and gathering of information, analysis, interpretation, and synthesis [4]. The multidirectional arrows represent intercon- nection between all the elements. Content analysis, observation, scenario writing and interviewing are some of the methods used for forecasting. Whereas content analysis is the most common approach preferred by the forecasters and it is more beneficial in long term forecasting. In today’s world, social networking sites such as Facebook, Twitter, Flickr, etc. are the most popular sources of social content data. It is available in various forms like numeric (e.g. number of tags or likes), text (e.g. comments), images (e.g. profile/timeline picture, posts), audio (e.g. songs), video (e.g. funny videos, trailers), etc. Image data is one of the most expressive forms of data. It contains interesting and useful information. Hence, image data is advantageous for content data analysis. Fashion trend analysis plays an important role in the fashion industries as it offers information about upcoming and outgoing styles. It is ultimately important while taking decisions in the business. Along with intuition, good judgment, creativity and analysis of social content data is helpful to get the facts about attributes influencing fashion trends. While task of analysis is very challenging due to its nature i.e. voluminous, unlabeled and heterogeneous. Various tasks of data mining such as classifica- tion, clustering, regression, association, outlier detection, etc. are useful for data analysis. According to the characteristics of social image data and the nature of problem, suitable data mining tasks are performed in the proposed system. https://doi.org/10.1016/j.asoc.2020.106517 1568-4946/© 2020 Elsevier B.V. All rights reserved.