Text Classification and Visualization on News Title Using Self Organizing Map

The presence of news articles in the world wide web (www) is growing very fast that give the challenge to analyze, organize, and label the documents. A method for analyzing document category is classification. However, the problems in text classification are visualization on the process and result. This study proposes the visualization of Self Organization Map (SOM) process for text classification of news title data. The visualization of topographic map of SOM may be used to gather new knowledge about the training process of neural network and enable the researcher to figure out the best parameter value for the SOM and also detect possible outlier data. From the numerical experiment, SOM proved reliable in classification with the highest accuracy 82.05%, and average 81.87% and it is better than the Naive Bayes method. SOM proved the capability to visualize the learning process and the results. SOM visualization can detect and show outliers of training data as a new knowledge.