In this paper, we have proposed a new modified forecasting method based on time-variant fuzzy time series. It uses trend heuristics in addition to high-order fuzzy logical relations and enhances the average forecasting accuracy significantly. To illustrate the whole forecasting process, we use actual enrollments (historical data for 22 years) of the University of Alabama (UA) and compare results obtained through other well-known fuzzy time series-based approaches described up to date in the literature. As a result, for all examined cases, the new time-variant method yields better forecasting accuracy as compared with alternative methods.