Product Sales Forecasting in Fashion Retailing

10 March 2022, 12:00 
zoom  
Product Sales Forecasting in Fashion Retailing

Morris Amara is a MSc student in the Department of Industrial Engineering
​Tel Aviv University​, 10 March 2022, 12:00 PM,  via zoom

 

Abstract:

This study outlines a methodology to forecast sales of fashion products in a retail environment by using a multi-dimensional recurrent neural network. It demonstrates how to achieve more accurate forecasts based on easy-to-collect signals originating from related products. It demonstrates how to build an rnn that includes these exogenous variables as informative exogenous variables that increase its forecasting power. It also addresses practical subject matters in the retail-fashion environment that are often overlooked in the literature, yet can significantly improve the forecasting results.
The proposed Multi-Dimensional Signal LSTM model, in short MDS-LSTM, can be applied to retail time series and is shown to significantly outperform the prediction accuracy of conventional retail forecasting methods, such as ARIMA and Seasonal ARIMA. The model was applied to thousands of real-world time series of a large international retailer. We show that the MDS-LSTM improves forecast accuracy by 6%-15%, leading to better customer experience and projected benefit of hundreds of millions of dollars.
The thesis includes an implementation of the proposed method in Python. The developed code was published as an open-source modular software, that can be utilized as a framework to expand and scale similar studies.

 

Bio:

Morris Amara is a MSc student in the department of Industrial Engineering. In the recent years, Morris specialized in implementing analytical solutions across global retail organizations. The research was conducted under the supervision of Prof. Irad ben Gal.

Tel Aviv University makes every effort to respect copyright. If you own copyright to the content contained
here and / or the use of such content is in your opinion infringing, Contact us as soon as possible >>