Abstract: Based on the principle of Bayes discriminant analysis,a Bayes discriminant model for evaluating expansive soil is established. Fiveindexes, including the liquid limit, the overall rate of expansion andcontraction, plastic index, moisture content, free expansion rate, areselected as the factors for synthetic evaluation of expansive soil. The shrink and expansion is divided into four grades as four normal populations in Bayes discriminant analysis. Bayesdiscriminant functions obtained through training a set of expansive soilsamples are used to compute the Bayes function values of the evaluatingsamples, and the maximal function value is used to judge which populationthe evaluating sample belongs to. The usefulness of the proposed model isverified by Jackknife method. The study shows that the average predictionaccuracy is 13.3%