Recently, we live with a huge amount of data. For example, we have great amount of news articles everyday. But there are small amount of useful information in the articles and it is hard to extract useful information manually. As a result, there are lots of news articles but, it is hard to read all of articles and find informative news manually. One of solutions on this problem is to summarize texts in the article. There are many studies on the text summarization techniques, but small number of studies to predict whether the article should be summarized or not. If we don’t know about that, it is likely to waste computing resources to summarize unnecessary articles. In this paper, we propose a method to model the pattern of user's summarization needs on news articles. We perform experiments using news articles and apply data mining techniques (C4.5 and Naïve Bayes) to model common preprocessor to execute the automatic summarization. Finally, we can get some meaningful results on the “desire to summarize” prediction.