A Study on SME Traditional Food Product Feedback by Leveraging Word2vec

The impact of reviews on sellers can be seen in improving services, products and communication. Small and medium-sized businesses (SMEs) sadly lack the ability to properly manage review data on social media channels, which might provide insightful analysis of the things they sell. This study focuses on examining the review associated with traditional food in e-commerce and find the accuracy of Skip-Gram and Continous Bag of Words (CBOW). The review will use word2vec as a word embedding. Word2vec will predicts words based on Skip-Gram and CBOW. For modeling data the experiment use random forest and support vector machine and evaluated by confusion matrix. The result presents that Continuous Bag of Words surpasses the accuracy of skip-gram in two conditions splitting data which is 83,5% for 9 training: 1 testing and 85,5 for 7 training: 3 testing.