Establishing trade occurrence on social media platforms

Abstract

Evasion of Value Added Tax (VAT) is a phenomenon that is contributing to substantial decrease of government revenues being collected across continents. One of the forms of VAT evasion cases is that of businesses failing to register with the tax authorities, inherently evading any tax liabilities. With the increase in popularity of social media, businesses have started using such platforms to advertise their products and services, and even building connections with other users. This allows businesses to achieve further reach and subsequently a potential increase in revenue. Naturally, people who engage in trade without having registered with the tax departments also feature on these social media platforms. The challenge is to identify all of the businesses amongst the rest of the users requires one to traverse millions of pages, making it humanly impractical. This dissertation proposes an automated tool which addresses this problem and targeting Maltese businesses. The tool is capable of streaming data from social media platforms, processing it and extracting useful features which help classify Maltese businesses. The tool incorporates a Naïve Bayes classifier which is trained using a supervised learning approach. The classifier identifies patterns within the given training dataset, allowing it to make informed decisions when it comes to predict unseen instances. The output generated by the classifier is then verified by a domain expert and necessary actions are taken. In conclusion, this tool provides a tangible solution to mitigate a significant portion of VAT evasion through Big Data analytics. This ultimately improves the state of the economy and the citizens would enjoy better benefits and services. Furthermore, the tool is maintainable and is modifiable to suite the different requirements of other countries and changes in economic activities.