In recent years, there has been an increased demand for future prediction in relation to social affairs, progress in science and technology, and economic circumstances. Moreover, there are large amounts of text data on the web, and there has been research into methods of future prediction support using natural language processing technology aimed at this. In previous research, we confirmed the effectiveness of future prediction supporting sentences, using an answer model generated through the use of pattern combination-based machine learning that considers language processing in word order for sentences referring to future events (FRS) using a newspaper corpus as learning data. However, there is other effective information on the Web in addition to just newspaper corpus, and this can also be used for sentence supporting prediction. Additionally, a method of improving prediction accuracy is to prepare an FRS classifier for each domain considering the sentence characteristics of each news domain, and acquire sentences supporting prediction. In this research, we obtained prediction support sentences related to future events from the news corpus on the Web, proposed a future event prediction support method using word polarity information, and showed that the prediction results exceeded the results of previous experiment.