Classification of Meme Sentiment Analysis On Kaggle.com Sites To Use Support Vector Machine Algorithm

Abstract

Meme is currently one of the media that is often used to convey a message or opinion on a topic that is currently hot in the community, and is widely discussed on social media. Apart from being a means of humor, memes are also commonly used as a medium to convey satire, even 'ridicule' to a party. This encourages curiosity to capture and classify memes circulating on social media, including through public data available on the Kaggle. This study aims to classify memes into three classes of sentiment, namely positive, neutral, and negative. In this case, the researcher uses Support Vector Machine algorithm with Radial Basis Function kernel because it can produce the highest accuracy compared to other kernels. The dataset downloaded through the Kaggle website is in the form of memes that have been labeled and accompanied by Optical Character Recognition (OCR) results consisting of a total of 6,992 meme data. By using Support Vector Machine algorithm, the classification results are obtained at 73.75% while using Naïve Bayes algorithm to obtain an accuracy of 61.24%. This proves that the application of Support Vector Machine algorithm in document classification is able to produce a fairly high accuracy when compared to the Naïve Bayes algorithm