The present invention provides a method for predicting the probability of criminal suspicion, which includes obtaining relevant information of the person to be tested, and determining the type of criminal record corresponding to the person to be tested according to the relevant information, and the relevant historical data screened out by the specified index corresponding to each type of criminal record; determining Detection time, and according to the relevant historical data, numerically assign values to the specified indicators in each criminal record type; set the feature vector to be formed by all the indicators in the specified indicators, and according to the feature vector and the assignment of the specified indicators, get each criminal record type of training sample library; the classification attributes of the training sample library of each type of criminal record are divided into 1 and 0, and the category probability when the classification attribute in the training sample library corresponding to each criminal record type is 1 is respectively adopted the logistic regression model Fitting is performed to obtain the criminal probability of each type of criminal record of the person to be tested. By implementing the invention, the crime type and crime probability of the person to be tested can be accurately predicted, and on-site guidance can be provided for key investigations by public security officers.