The development of a successful short-range forecasting model for predicting enrollment in a university intensive English course began with examination of the characteristics (nationality, enrollment patterns, persistence) of both course participants and no-shows; the resulting figures revealed consistent patterns for country of origin and for former versus new students. Show-up rates were derived from these patterns, and several prediction models were applied to registration data for several semesters. This analysis revealed further patterns in walk-in registrations and in the differential timing of registration according to the student's country of origin. A database was developed to record, at the time of application, the information found to be relevant in these analyses, including name, country of origin, geographic area, semester applied for, date of application, new- or former-student status, and current location (in the United States or abroad). The results for each semester are incorporated into the next semester's prediction formula. Other impact factors, such as a country's changing educational or political policy and economic situation, changes in U.S. immigration regulations, and changes in university policy are also considered. These considerations are quantified when possible and integrated into the forecasting model. An initial forecast is made ten weeks before the semester begins and is revised weekly until classes begin. An appended handout section presents supporting data including graphs, forecasting formulas, and an enrollment forecast table for the past five semesters. (MSE)