Text classification using Fuzzy TF-IDF and Machine Learning Models
Mariem Bounabi, Karim El Moutaouakil, Khalid Satori · 2019
The representation of the information has an important impact on the text classification task. Several weighting methods were proposed in the literature, and the term frequency-inverse term frequency (TFIDF), the most know on the text treatment field. The FTF-IDF is a vector representation where the components of the TFIDF are presented as inputs to the Fuzzy Inference System (FIS). In this work, we compare several Machin Learning algorithms such as Naïve Bayes and its derivatives, SVM and Random forest classifiers, using the FTF-IDF representation. To improve the quality of the used classifiers, we call sum attribute selection methods. The recognition rate, for the tested systems, is satisfied, where the system based on naïve Bayes classifier, the FTF-IDF weighting terms, and the info gain select attributes method gives 98.7% as accuracy.