摘要：In this modern era of digitization, the competition is significantly increasing among retailers. One of the major challenges for them is demand prediction or sales forecasting. Especially in this Covid pandemic, retail sales forecasting became very crucial due to the employee shortage, and increasing online demand. In the modern era of digitization, competition is increasing. This research explores the application of an advanced deep learning approach in predicting the market demands in advance of individual products for the future seasons. This application aims to support an American Multinational Retail company in ordering, purchasing, and managing inventory. Accordingly, the company provides a real sales dataset to perform this study. This research proposes two sales forecasting strategies based on LSTM and LGBM models. We first execute data preprocessing techniques using statistical feature engineering on the raw sales data. Thereafter perform the LSTM and LGBM algorithms for training and prediction. LGBM takes past data from lag feature engineering for better forecasting. For that, we found that LGBM performs better than LSTM in forecasting.
