Ranking of branded products using aspect-oriented sentiment analysis and ensembled multiple criteria decision-making

Abstract

In this paper, we proposed a novel aspect-oriented sentiment analysis and combined it with an ensemble of well-known multiple criteria decision-making (MCDM) methods such as TOPSIS, VIKOR, PROMETHEE II, ELECTRE III and FMADM for comparing a given set of products and services. To demonstrate the effectiveness of the proposed approach, we performed comparative study of a set of mobile phones and digital cameras using customer feedback available on twitter and e-commerce customer reviews. Eventually, the final ranks of the products were obtained by ensembling five MCDM methods through simple majority voting. During the whole process, we contributed to the state-of-the-art into four folds: first, we developed a novel aspect-oriented opinion determination approach. Second, proposal of a hybrid aspect-level sentiment score computation approach. Third, we developed two domain ontologies in order to represent prominent aspects and their synonyms for mobile phone and digital camera. Finally, we combined aspect-oriented sentiment analysis with ensembled MCDM to perform aspect-level comparative study of a set of alternatives. Further, VIKOR sensitivity analysis is also performed. Eventually, a detailed visual analytics is presented on a geometrical analysis for interactive aid (GAIA) plane.