Authors: Etienne DAVID, Jean BELLOT and Sylvain LE CORFF

Paper link: https://arxiv.org/abs/2202.03224

Developing models and algorithms to draw causal inference for time series is a long standing statistical problem. It is crucial for many applications, in particular for fashion or retail industries, to make optimal inventory decisions and avoid massive wastes. By tracking thousands of fashion trends on social media with state-of-the-art computer vision approaches, we propose a new model for fashion time series forecasting. Our contribution is twofold. We first provide publicly an appealing fashion dataset gathering 10000 weekly fashion time series. As influence dynamics are the key of emerging trend detection, we associate with each time series an external weak signal representing behaviours of influencers. Secondly, to leverage such a complex and rich dataset, we propose a new hybrid forecasting model. Our approach combines per-time-series parametric models with seasonal components and a global recurrent neural network to include sporadic external signals. This hybrid model provides state-of-the-art results on the proposed fashion dataset, on the weekly time series of the M4 competition, and illustrates the benefit of the contribution of external weak signals.

This repository provides the code of the HERMES approach and a simple code base to reproduce the results presented in the paper. The repository is organized as follow:

• model/: Directory gathering the code of HERMES
• run/: Directory containing a script to reproduce the results of the paper.
• data/: Directory gathering the fashion dataset presented in the paper.
• docker/: directory gathering the code to reproduce a docker so as to recover the exact result of the paper.

First, you should build, run and enter into the docker. In the main folder, run

make build run enter

To reproduce the result of HERMES on the sample of 100 time series:

• run_hermes.py run

python run/run_hermes.py --help # display the default parameters and their description
python run/run_hermes.py --model_dir_tag hermes_100ts --nb_time_series 100 --rnn_lr 0.0005 --batch_size 8 --load_pretrain_stat_model # train 10 hermes with different seeds on the sample of 100 fashion time series and save the results in the dir result/
python run/run_hermes.py --model_dir_tag hermes_1000ts --nb_time_series 1000 --rnn_lr 0.001 --batch_size 64 --load_pretrain_stat_model # train 10 hermes with different seeds on the sample of 1000 fashion time series and save the results in the dir result/
python run/run_hermes.py --model_dir_tag hermes_10000ts --nb_time_series 10000 --rnn_lr 0.001 --batch_size 64 --load_pretrain_stat_model # train 10 hermes with different seeds on the whole fashion time series and save the results in the dir result/

The following tabs summarize results that can be reproduced with this code:

• Experience on 10000 time series:

• Experience on 1000 time series:

• Experience on 100 time series: