Improving sentiment reviews classification performance using support vector machine-fuzzy matching algorithm

High dimensionality in data sets is one of the challenges faced classification, mining, and sentiment analysis. In set, many dimensionalities require effort to simplify. Many these have a major impact on complexity performance algorithms used for classification. Various were encountered, including how determine optimal combination pre-processing techniques, clean dataset, best classification algorithm. This study uses new approach based three powerful techniques which are: tokenizing-lowercasing-stemming (for series preprocessing), support vector machine (SVM) supervised fuzzy matching (FM) reduction. The proposed model was realized using 3 different datasets, namely Amazon product review, movie airline review from Twitter. provides better findings than previous results. Improved generated by SVM combined with FM, resulting 96% accuracy. So that SVM-FM can be said analysis given set.