In the increasingly competitive fashion retail industry, companies are constantly adopting strategies focused on adjusting the products characteristics to closely satisfy customers' requirements and preferences. Although the lifecycles of fashion products are very short, the definition of inventory and purchasing strategies can be supported by the large amounts of historical data which are collected and stored in companies' databases. This study explores the use of a deep learning approach to forecast sales in fashion industry, predicting the sales of new individual products in future seasons. This study aims to support a fashion retail company in its purchasing operations and consequently the dataset under analysis is a real dataset provided by this company. The models were developed considering a wide and diverse set of variables, namely products' physical characteristics and the opinion of domain experts. Furthermore, this study compares the sales predictions obtained with the deep learning approach with those obtained with a set of shallow techniques, i.e. Decision Trees, Random Forest, Support Vector Regression, Artificial Neural Networks and Linear Regression. The model employing deep learning was found to have good performance to predict sales in fashion retail market, however for part of the evaluation metrics considered, it does not perform significantly better than some of the shallow techniques, namely Random Forest.
Fashion retail is a highly competitive market, where inventory control plays a key role in the profitability of companies. Accurate sales forecasting is therefore fundamental to be successful in this environment. If sales forecasting is not accurate, stock-out or overstock situations might occur, which can have a direct and immediate impact on the company’s profitability (Agrawal & Schorling, 1996; Sun, Choi, Au, & Yu, 2008; Baecke, De Baets, & Vanderheyden, 2017; Sodero & Rabinovich, 2017; Xia & Wong, 2014) . The effect is not restricted to profitability performance, as quality of the customer service can also be affected by an inefficient forecasting system. For example, if a customer is faced with a stock-out situation, they might decide to shop in a different retailer (Corsten & Gruen, 2004). Additionally, it is known that the fashion industry operates with long supply chains involving a large number of actors (such as raw materials suppliers, manufacturers, distributors and retailers), which leads to orders being placed before there is an accurate understanding of the demand level for the products (H. L. Lee, Padmanabhan, & Whang, 1997; H. Huang & Liu, 2017). Despite its relevance, sales forecasting is a complex subject because the sales’ success of a product is highly dependent on the personal taste of consumers, which varies greatly (Allenby, Jen, & Leone, 1996; Choi, Hui, Liu, Ng, & Yu, 2014). In addition, the lifecycle of fashion products is typically very short, being replaced every new season by new products with no historical sales data (Choi, Hui, Ng, & Yu, 2012). Fashion collections are also composed of an extremely large number of different products in many different sizes, corresponding to many different stock keeping units (SKUs) (Liu, Ren, Choi, Hui, & Ng, 2013). Moreover, several external factors can also have a direct impact on the sales including the weather conditions, holidays, marketing actions, promotions, fashion trends and the current economic environment (Sébastien Thomassey & Fiordaliso, 2006; Sun et al., 2008; Ni & Fan, 2011).