Effective text classification using multi-level fuzzy neural network

Nowadays, large volumes of text data are being produced in real time due to expansion of communication. It is necessary to organize this data for exploitation and extraction of useful information. Text classification based on the topic is one of the efficient solutions to this problem. Efficient algorithms are applied for text classification if they address high dimensional data. In this paper, a novel neural network classifier is adapted to classify the texts and compared to some similar techniques. To this end, firstly preprocessing is conducted on the words, which includes five steps. After that, the feature extraction phase is applied consisting of three steps, with the Principle Component Analysis (PCA) method used for dimensionality reduction. Finally, a supervised learning method, namely the Multi-Level Fuzzy min-max neural network classifier (MLF), is used for text classification. The simulation results of the experiments, compared to three other principle methods, indicate that MLF method has a high accuracy of 94% for the Reuters-21578 dataset and 95% for the 20 newsgroup dataset. In addition, the execution time of the proposed method is less than other supervised learning algorithms.