Approximately 40%–65% of new fitness club (FC) members cancel their membership within 6 months. To prevent such cancellations, it is essential to identify members at high risk of doing so. This study developed a model to predict the probability of discontinuing FC membership among new members. We conducted a cohort study and enrolled participants from 17 FCs in Japan. We asked 5,421 individuals who became members from March 29, 2015 to April 5, 2016 to participate in the study; 2,934 completed the baseline survey, which was conducted when the participants became FC members. We followed up the participants until September 30, 2016. We excluded 883 participants with missing values and 69 participants under aged 18 years; thus, our analysis covered 1,982 individuals. We conducted the random survival forest to develop the prediction model. The mean follow-up period was 296.3 (standard deviation, 127.3) days; 488 participants (24.6%) cancelled their membership during the follow-up. The prediction model comprised 8 predictors: age; month of joining FC; years of education; being under medical follow-up; reasons for joining FC (health improvement, relaxation); and perceived benefits from exercise (maintaining good body weight, recognition of one’s ability by other). The discrimination and calibration were acceptable (C statistic: 0.692, continuous ranked probability score: 0.134). Our findings suggest that the prediction model could assess the valid probability for early FC cancellation among new members; however, a validation study will be needed.