Clustering Text Data Using Text ART Neural Network

Most studies of data mining have focus on structured data such as relational, transactional, and data warehouse data. However, the most available information is stored in text database, which consist of large amounts of text documents such as news articles, research papers, and e-mail messages. Data stored in most text databases are unstructured data, such as abstract and contents. The ability to deal with different types of attributes is a typical requirement of clustering in data mining. Thus, mining unstructured data has become an increasingly important task in text mining. The main contribution of this paper is to cluster on a data set, which has a non-numerical feature value. Unlike the conventional clustering algorithms such as the K-Means algorithm, which forms clusters in numerical values domains, a Text ART Neural Network works directly on textual information without text transformation into a numerical value. The experimental results are represented that conducted on 2 datasets. The first dataset is a Synthesized Text Document and the second dataset is a Reuter-21578 Distribution 1.0. The F-Measure equation use to measure the effectiveness of the proposed technique. According to the experimental results, the proposed neural network has well performance in clustering text data that the F-Measure of the experiment is 95.56% and 83.31% respectively. Key-Words: Text Mining, Document Clustering, Unsupervised Learning, Artificial Neural Networks