The Question Answering Systems (QASs) use method of information retrieval and Information extraction to retrieves documents that contain special answers to the question. One of the existence problems is finding the desired information from this very high variety. For this reason, it is necessary to find ways for organizing, classification and retrieving of information. Question classification plays an important role in providing a correct answer on QASs because giving a bunch of formulated questions to provide the correct answer from among the many documents will be highly effective. The aim of classification is selecting suitable label for questions based on the expected response. In this paper, we investigate the effect of automatically classifying questions on machine learning algorithms. In this paper, we will explain different types of algorithms and compare and evaluate them and next we will investigate the existence algorithms' weakness and advantage in question classification. As a result, in the past most classification was done based on sets of words that many studies show that to maximize the efficiency of the classification of algorithms we require semantics and in the questions we should looking for feature that be close to the meaning of questions. A great deal of research proposed to analysis and to classify emotions and to extract knowledge from them and to classify them using semantic and linguistic knowledge but it still requires a lot of research and development.