English preposition error correction system is to help English language learners to correct automatically the common mistakes of English prepositions. First, to classsify and count up the preposition errors in the marked corpus files, sum up 21 kinds of common prepositions, with English wiki corpus and use of computer algorithms automatically error interpolation algorithm to get the training set, and then based on the training set, by using a classification based on the maximum entropy model chosen, including context, prepositions complement other features, training model on the training set, and then use the model to predict the input sentence and correct use of the presence of errors. In NUCLE corpus experiment, given corpus processing, model features, size, number of iterations to test the effect of the impact of training data set, and compare the results of the Naive Bayes model, and finally to the F value 27.68 in the test data with respect to the shared task CoNLL2013 best results have slightly improved.