Sentiment Analysis for Seller Integrity Authentication on a Business Page

Abstract

Markets have transformed over time from being predominantly analogue to becoming increasingly digital,presenting a plethora of opportunities for businesses to exploit in terms of improving and streamlining business processes. As technology advanced, the digital markets have given businesses the platform for better visibility and more stability. However, a buyer's inability to accurately evaluate sellers’ authenticity, based on ratings and reviews on items they wish to purchase on a business page, presents a significant vulnerability for dubious sellers who may use fraudulent ratings and reviews to exploit unsuspecting buyers. This study adopts a machine learning approach to authenticate seller integrity on business pages through sentiment analysis. Data was extracted from customer reviews and ratings from selected brands of digital products across five major ecommerce sites: Jumia, Konga, Amazon, eBay, and Aliexpress. The model was built using Support Vector Machine algorithm to categorize the sentiments expressed in reviews as positive, negative, or neutral. The machine learning approach was selected due to its effectiveness in pattern recognition, adaptability to evolving data patterns and its suitability in providing high accuracy in sentiment classification. The model's performance was assessed with the training dataset yielding 99.58% accuracy and the test dataset achieving 97.27% accuracy. The results present a reliable method for enhancing consumer trust in online marketplaces by verifying seller authenticity based on their ratings and reviews on a business page.