OPINION MINING TERHADAP RESENSI VIDEO GAMES MENGGUNAKAN SUPPORT VECTOR MACHINE; OPINION MINING OF VIDEO GAMES REVIEW USING SUPPORT VECTOR MACHINE

Setiawan, Dhiaz A, Aina Musdholifah

2015 | Skripsi | FMIPA UGM

Currently the video games industry is one of the rapidly growing industry with tens to hundreds of titles released each year. However, not every title has the same quality and satisfaction leading many people expressing their thought on a review which others can read and used as a consideration in determining which title to be played. Opinion is a task of data mining that can be used to extract the opinions of those reviews. This research is trying to build a system that can do opinion mining on video games reviews using Support Vector Machine (SVM) as a classifier. This research also comparing the commonly used kernels, such as linear kernel, polynomial kernel and RBF kernel. In addition, the number of attributes used on each review or instance is limited to 500, 1000 and 1500 using information gain algorithm. The results of each respective attribute number will also be compared. There are two types of testing method in this research, which are 10-fold cross validation test and testing using an independent testing dataset. Both of the tests show that there are no much differences in using different kernels and using different amounts of attributes, which are only amounted to 0-2%. Using 1000 attributes resulted in highest accuracy on both tests, which are 89.68% on 10-fold cross validation test and 81.72% on test using test set. Those values are achieved using third degree polynomial kernel on 10-fold cross validation test and using linear kernel on test using test set.

Kata Kunci : opinion mining; data mining; support vector machine; video game review.