FASHION TREND FORCASTS BASED ON TIME SERIES

Aiming at the problem that the traditional fashion trend forecasting methods is inefficient and highly depends on the subjective will of experts and users, so the training data is difficult to reflect the real fashion trend, we propose a model for predicting fashion trends based on LSTM and fashion week image information. This method crawled the show pictures of the four major fashion weeks from 2013 to 2021 in the fashion website vogue. The picture information was analyzed and the show picture information was combined with internal fashion knowledge. The LSTM model based the attention mechanism was used to find fashion relationships from time-series to predict fashion trends. The experimental results show that this method performs best on multiple data sets.