Sentiment Analysis to Support Marketing Decision Making Process: A Hybrid Model

Marketers aim to understand what influences people’s decisions when purchasing products and services, which has been proven to be based on natural instincts that drive humans to follow the behavior of others. Thus, this research is investigating the use of sentiment analysis techniques and proposes a hybrid approach that combines lexicon-based and machine learning-based approaches to analyze customers’ review a major e-commerce platform. The lexicon approach was firstly applied at a word-level to explore the reviews and provide some preliminary results about the most frequent words used in the reviews in a form of word-clouds. Then, the lexicon approach was applied to sentence-level to obtain sentiment polarity results, which was used to train machine learning models. Next, the trained models were tested on un-labelled reviews (test data); proving that Naive Bayes (NB) outperformed other classifiers. The hybrid model described in this research can offer organizations a better understanding of customers’ attitudes towards their products.