Comparison Study Of Term Weighting Optimally With SVM In Sentiment Analysis

Abstract

The rapid of internet and social media users have changed the way people interact in their daily activities. For example, banking and retail began to use various social media, especially online media such as tweeter. The problem that arises is how to get information from thousands and even million data generated through social media, to be a decision as in predicting consumer satisfaction of the service or product.Another problem is the social media users in communicating using slang or local language. In sentiment analysis to predict the sentiment is not easy because it must be able to identify the words. In sentiment analysis, to overcome these problems the method used is text mining so as to process opinions from social media. The proposed approach is to analyze optimal term weighting between TF-IDF, frequency term (TF) and Binary Term Occurrence (BTO), using SVM algorithm. Target feature extraction for selection of datasets by predicting positive and negative sentiments. The result of weighting of terms approaching sentiment is using TF-IDFwith SVM.

A study of the possibilities of text mining and machine learning for score evaluation and review content

“…We used Rapid Miner Studio ver9.6 for text mining and machine learning for labeling. [11][12][13][14][15] Text mining was carried out in combination with Text Processing, an extension of Rapid Miner Studio.…”