PERANCANGAN SISTEM ANALISIS SENTIMEN UNTUK KLASIFIKASI SENTIMEN KONSUMEN TIKET.COM DI TWITTER

• Main Author: Hanifah Syarief, Aqila
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/66092
• Institution: Institut Teknologi Bandung

Machine learning helds great role for companies to implement data driven design that is important for increasing customer loyalty. tiket.com is an Online Travel Agent (OTA) with vision to be the most loved Online Travel Agent (OTA) & lifestyle app. However, implementation of machine learning in the form of text mining to increase company’s ability on data driven design hasn’t been facilitated. Therefore, this research aims to create a system for the company to analyze market’s sentiment from tiket.com’s related tweets. This research has two parts which are model creation and prototype creation. Model creation was done with two algorithms – Support Vector Machine and Naïve Bayes. 22,595 data was preprocessed and used to build the model. Model creation and model evaluation was done by different data. This data comes from splitting the initial data with proportion of 70:30 for train data and test data respectively. The classification that’s implemented in this research divides the data into three classes – Good, Bad, and Neutral. Model that was built with algorithm of Support Vector Machine was selected as the best model with accuracy of 81,96%. Model with Naïve Bayes algorithms has accuracy of 78,26%. The first model was chosen to be implemented in the prototype that’s built as an application to predict future data’s sentiment. On the prototype, several interfaces were shown in the form of pie chart, word cloud, and table as per user’s requirement.