Classification of user-level twitter polarity using soft computing approach

Twitter has become one of the most commonly used Social Networking Services in modern society, especially among young people. More and more users choose to express their opinions on and share their emotions on general public instead of posting the words on forums or keeping the words in dairy. So, tweets are very good samples for the user’s sentiment analysis. In this project, I aimed to finish some simple examples of sentiment analysis, classification of polarity and content visualization by applying Twitter's Search and Streaming APIs. I choose Matlab to perform the sentiment analysis. The experiment will be conducted between the tweets which contain two selected competing brands. The sentiment polarity of tweets will be classified as 3 levels: positive, negative and neutral. There are four code section: one section to retrieve raw tweets which are contains the keyword for user’s expectation; one section to clean the raw data, extract the important subjective words, convert the data to structured array and calculate the sentiment score; one section to load the input array, classify the polarity level, calculate the polarity distribution, NSR and plot the graph; one to realize the content visualization. After analyzing results, it is convenient to make a clear comparison between input competing brands’ reputation from different figure and table. There is also a figure to summarize the high frequency words that appear in retrieved tweets. Project No: P1019-172 There are still some limitations in this project, I will cover the suggested solutions in this report and hope to solve these problems in the follow-up research.Bachelor of Engineering (Electrical and Electronic Engineering