Abstract
In the fashion industry, fads, a phenomenon in which demand for specific items surges in a short time and then disappears, are often observed. However, the conventional demand forecasting methods have limitations in predicting fads because it focuses on long-term trends. This paper presents a new approach called Fad2Vec, which adapts the Item2Vec approach to detect fads effectively. Hot periods composed of a hot point and a preceding period for each product are identified based on time-series sales data. Products are embedded into vectors representing the time points when they are temporarily fashionable. Then, they are clustered into a fad group where fashion items belong to the same cluster and show similar fad patterns. A case study of four famous fashion brands is provided to show how Fad2Vec works and verify its validity. The proposed Fad2Vec approach is expected to be practically utilized for the production planning of fast fashion companies.
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
