To solve the problem of hard employment of graduates who expect for the impractical emolument, the paper builds a model for emolument prediction. On the basis of an classification algorithm with natural neighbor(NaN)，it analyzes the employment data of graduates majoring in Information Engineering in past three years. The paper uses factor analysis method to fetch the latency of employment emolument level determinants. Classification predicts the emolument by applying the latency as a variable based on the classification algorithm. This algorithm avoids the difficulty of parameter selection in K-nearest neighbor(KNN). The neighbors of each node can also be acquired as the topography of data set. According to the experiments, the prediction accuracy is 80.16%. The paper can guide graduates to build a reasonable emolument prediction or improve employment.