Text classification is one of the natural language processing (NLP) methods that assigns texts to one or more categories. In this paper, we propose a text classification method based on deep neural networks and word clustering. We also provide analysis on the effects of the number of channels, the way of converting the cluster information to a vector, and the update method of the input channel during learning with baseline. To show the effectiveness of our approach, we apply the method to the TREC question dataset and Movie Review dataset. From the results, we confirm that semantic features from word clustering is able to increase the classification accuracy by 1.96%.