Social Forecasting Tools forConsumer Goods Personalisation

Abstract

Since the rise of WEB 2.0, social networks, blogging and microblogging sites, an abundance of user-generated content (UGC) is produced daily. These can be a goldmine of information when utilized for modelling trends and analysis, as future trends could be predicted once similar UGC is generated in the future. This project aims to build a machine learning (ML) forecasting tool by modelling social media’s users’ metadata to predict general consumers’ purchasing trends for fast-moving consumer goods (FMCG). This will use opinion mining, sentiment analysis and sentiment classification tools to capture the overall public sentiment towards products that have undergone price, quality, packaging, and flavour changes, or been offered under promotions, and comparing these with actual historical demands for the products under study. The main source of data collected will be the microblogging website Twitter and Amazon Marketplace. According to a joint study by Twitter and the analytics firm Annalect, close to 40% of the study correspondents said they have purchased an item online after seeing it used on social media outlet (Tankovska, 2021). Moreover, Twitter is the only true “open data” online source that is open to the public (Petrov, 2021), with 192 million daily active users, producing on average 500 million tweets per day (Chae, 2014).