PURPOSE: Based on Transtheoretical Model, exercise stages (ES) are associated with processes of change (POC), which are strategies that individuals use to change and improve their exercise level, such as enlisting social support or setting a goal for exercise. But how the POC would classify exercise levels remains unaddressed. This study examined this unclear aspect. METHODS: Exercise Stages of Change survey and POC survey (Marcus & Forsyth, 2009) were administered to 161 university students in the US. The former survey classified participants into five ESs: No participant was identified in Stage 1 (precontemplation): no exercise and no intention to change; 27 were in stage 2 (S2, contemplation): no exercise but considering a change; 27 in stage 3 (S3, preparation): having some change/exercise; 49 in stage 4 (S4, action): performing regular exercise recently; and 58 in stage 5 (S5, maintenance): performing regular exercise as a habit. The POC survey produced 10 predictors (i.e., 10 exercise promotion strategies). Discriminant analysis was run to examine how accurately the predictors could classify the exercise stages for participants. Further, kappa was computed to assess classification accuracy as well for correcting chance agreement. RESULTS: “Substituting alternatives” (i.e., replacing sedentary time with some type of PA) was the only identified significant discriminant function (DF) among the 10 predictors with its DF coefficient = .79 and discriminant loading = .84, and the significant tests showed Λ = .40, χ2(30) = 142.08, and p < .001. The eigenvalue of this DF = 1.36, and its % of variance = 94.9, accounting for 94.9% of the discriminating ability of the predictors. As for classification accuracy, 74.1%, 37.0%, 40.8%, 72.4%, and 57.1% of participants were correctly classified in S2, S3, S4, S5, and all Stages, respectively. Additionally, a kappa value of .40 showed a moderate agreement in classification. CONCLUSIONS: “Substituting alternatives” is the only significant DF of the predictors and occupies 94.5% of the total discriminating ability in differentiating the exercise levels. It makes sense that acting to exercise is the major aspect differentiating the exercise levels. The predictors, as a whole, demonstrate a moderate classification function, with 57.1% of overall classification accuracy.