Analytics For An Online Retailer: Social Media Big Data Analytics For Sales Forecasting

Abstract

Demand forecasting has been a common topic for many years, especially after COVID 19, the search for solutions to uncertainty has been more focused than ever. Social media has proven its place in other fields such as travel and entertainment when it comes to forecasting, yet it seems to be overlooked by academia in the field of logistics. This thesis aims to prove the promising potential of social-media-based data in forecasting demand in the e-commerce field. The data included in this study is drawn from a Vietnamese cosmetic company on the most well-known e-commerce platform in Vietnam - Shopee. Datasets combine (1) baseline data which includes sales, prices, rating, rating records, and discount programs; (2) social-media-based data which includes information related to likes, comments and shares on Facebook. I implemented two most effective machine learning methods in solving two aspects of this study namely for social-media-based forecasting by Random forests model and e-commerce forecasting by an Advanced ARIMA and LSTM model.