Application of Multilayer Perceptron for Digital Society Sentiment Analysis towards Indonesia Biggest E-commerce Platforms
Kevin Kelly Isyanta · International Journal of Advanced Trends in Computer Science and Engineering · 2020
Indonesia is one from many countries in the world which has the highest growth of e-commerce platforms.By the mean of dynamic website technology, through their social media accounts, digital societies often express their opinion on e-commerce platforms.Oftentimes, on a large scale, capital leading e-commerce companies try to evaluate sentiment polarity of their new promotion or sale strategy through such data.However, as a copious amount of opinions from e-commerce's various social media accounts, the manual labelling process might devour a large number of resources and prone to human errors.For tackling the aforementioned problem, previous researches show a promising result by encoding such problem into a machine learning classification task.Henceforth, in this research, we try to utilize multilayer perceptron (MLP), as one of machine learning classification algorithm, for tackling such problem and analyze the performance of it.In this research, we also present the effect of a comparatively new word embedding model, fastText and aTF-IDF as a well-known feature extraction method for MLP method.Our experiment shows that the developed MLP model able to achieves an accuracy score of 89.24% and F1-score of 89%.