Sentiment Analysis Product using Text Mining with Word Relationship Represented by Graph
Watchareewan Jitsakul · 2023
This paper presents a combination of text classifier and word centrality for analyzing product reviews. The product reviews collected from the Amazon website consist of 10,374 office product reviews. The classifier selected Random Forest algorithm and word centrality measures are Degree Centrality (DC), Centrality (BC), Closeness Centrality (CC), and Eigenvector Centrality (EC) and set threshold values ranking from 0.2-1.0. The result showed that Random Forest algorithm combining the world centrality was higher than the Random Forest alone with a threshold greater than 0.2. Classification accuracy is 85.50% to 87.50%.