Stemming has commonly used in some researchabout text mining, information retrieval, and natural languageprocessing. However, there is an indication that stemming does notdeliver significant influence toward accuracy in text classification.Hence, this research attempts to investigate the influence of thestemming process on Indonesian tweet sentiment analysis.Furthermore, this work examines about the difference effectbetween two conditions by involving stemming and withoutinvolving stemming on pre-preprocessing task. The experimentsshow that the accuracy difference for SVM using stemming in preprocessingacquired 0.67% and 1.34% higher than pre-processingwithout stemming, whereas, Naive Bayes obtained 0.23% and1.12%. Finally, this research proves that stemming does not raisethe accuracy either using SVM or Naive Bayes algorithm
Copyrights © 2015