ABSTRACT

here is a huge amount of data in various formats that is present over internet. It is difficult to classify millions of text documents manually because it requires more time and resources. Therefore, text classification is widely used for organizing text automatically. In this research, two classification techniques Naive Bayes Multinomial Updateable (NBMU) and Sequential Minimal Optimization (SMO) were applied on dataset. According to results, it was observed that, the combination of Rainbow Stopword (R) and Snowball Stemmer (SS) in NMBU classifier yielded the maximum accuracy (83%) while taking nominal time (0.07 sec) compared to other combinations of stemmers and stopwords removal. Whereas the SMO classifier yielded high accuracy (80%) by three different combinations of stemmers and stopwords removers, (Wordsfromfile) Stopword and Lovins Stemmer (WFF_LS), Regexpfromfile Stopword and Lovins Stemmer (REFF_LS),Regexpfromfile Stopword and Snowball Stemmer (REFF_SS)). However, the time taken for building the model was significantly high (500 - 1000 times higher). Based on the results of this research, it is suggested that the R & SS combination of stopwords remover and stemmer, respectively, in NMBU classifier perform best across the other selected combinations in terms of accuracy. By analysing the results, it is observed that the overall performance of SMO classifier in terms of accuracy is quite high on average compared to NBMU classifier. It was noted that the time that was taken by NBMU classifier was significantly low, compared to SMO classifier. Despite that, SMO is suggested to be utilised for the text classification due to the fact that the overall performance of this classifier is significantly higher in term of accuracy and that the text classification is a difficult task to perform, therefore, the difference in the time taken by NBMU and SMO become negligible and compensated by better accuracy.