The invention discloses a potato late blight occurrence prediction model based on meteorological conditions in northern China, including extraction of potato late blight disease and meteorological data set in northern China, potato late blight disease and non-disease sample set, the establishment of training set and test set, the construction of classifiers, determination of the parameter thresholds, and obtaining the results of the regional autoregressive and test. By analyzing the meteorological data of the nearby weather station in the monitoring area, based on the pathogenesis of late blight, disease samples are analyzed, the combination of meteorological condition variables and parameter thresholds affecting late blight are determined, and the overall accuracy of the model can reach about 78%. When the variable threshold is selected by ROC curve, the probability of false positive class is reduced as much as possible on the basis of ensuring the accuracy of the model, thus reducingthe problem of excessive use of pesticides in the environment and agricultural products caused by the occurrence of pesticides caused by the model error prediction.