SENTIMENT DYNAMICS ANALYSIS USING NAÏVE BAYES ALGORTIHM, ISING MODEL, AND CELLULAR AUTOMATA MODEL

• Main Author: Marwah Putri, Adnin
• Format: Final Project
• Language: Indonesia
• Subjects: Fisika
• Online Access: https://digilib.itb.ac.id/gdl/view/33661
• Institution: Institut Teknologi Bandung

Sentiment is an opinion or a view based on feelings and emotions towards some topics. Interactions that occur between individuals will lead to the dynamics of sentiment where a sentiment of an individual will influences the sentiments of others nearby individuals and this phenomenon can be modelled with physics. Complex System is a system that consists of some complex interacting agents and are able to form macroscopic patterns of the interactions of individuals with temporal, spatial or functional structures to form new collective behaviours. Sentiment analysis will be carried out in stages from text mining (the process of extracting information in the form of text data sources) using Rapidminer, classification based on positive / negative sentiments with machine learning of Naive Bayes (algorithm by determining the largest conditional probability value) using Tweetment, as well as some modeling using the Ising Model (phase transition modeling methods for ferromagnetism), and Cellular Automata (homogeneous cell lattice modeling with discrete conditions and transitions) using Matlab and NetLogo, then data processing using excel. In this Final Project, the author will process, analyze, and visualize public sentiment data from social media (twitter) regarding online stores within a period of 2 weeks so that the results related to the positive sentiment from Amazon, Alibaba, eBay, and TaoBao is obtained and will be sorted by its level. In addition, it was concluded that the factors of the interactions nearby, factors of extreme events, and factors of majority opinion will influence the emergence of changes in sentiment towards time (dynamics of sentiment). Also, the ratio of the difference in sentiment factors, the rule of interaction factors, and number of population factors will influence the time needed to reach consensus, a situation in which the dynamics of sentiment is stable enough and sentiment is no longer excessively changing.