Fashion Colour Predictions by Social Media

Precise estimation of order volume is a critical factor affecting apparel company profit. An accurate order volume could reduce the chance of over-stock and out-of-stock. Social media has been identified its importance in influencing consumer behavior. The proposing project will study the effects of social medial impacts on apparel fashion color. The project would develop a mathematic model, to predict garment sales volume precisely and then place production order quantity accordingly in different colors, and reduce the over-stock percentage. Investigation on the impact of social media data on color fashion preferences would be conducted. Social media data would be collected from facebook.com and weibo.com in English, Traditional Chinese, and Simplified Chinese, focusing on Hong Kong and Mainland China users. Validation of the model would be conducted using real-time social media data and frontline sales data (i.e. sale volume, stock volume, and price etc.) from at least one apparel company on a rolling base in the project period. A/B test would be conducted to measure the performance of the model and compare with the apparel company’s original in-house prediction method. The outcome of this project will be a tool helping company estimate manufacture order volume.
Application

The model makes use of a database comprising millions of posts from Facebook and Weibo, two of the most popular social media sites in the Greater-China region. Posts in both English and Chinese are used in the model. Authentic fashion posts that relate to colour are identified through Natural Language Processing (NLP). The model also applies advanced machine-learning methods to improve the accuracy of fashion colour prediction.

The project studied the transmission pattern of fashion colour information in social media and generated equations based on posts from fashion brands, magazines, designers and key opinion leaders.

The model can be customised for different users, based on their market positions and production lag time. In other words, the tool can be modified to match appropriately each user’s particular features and needs.

Industry Benefit

Upgrading traditional experience-based decision-making with analysis of current big data and machine learning modeling techniques, this project has great potential to generate fashion trend estimations in terms of colour range, style, fitting, and may even function to help drive a form of “new retail”, both on- and off-line.

Awards
46th International Exhibition of Inventions of Geneva (2018) - Gold Medal & Special Prize from National Research Council of Thailand
Project Name (ITF)
Impact of Social Media on Colour Fashion Preferences
Project Number (ITF)
ITP/034/16TP