Over the last few years, a large number of Fuzzy Time Series (FTS) forecasting models have been formulated and proposed in the literature to handle the complex and incomplete problems. However, the accuracy of a model is problem specific and varies among data sets. Though numerous models claimed superior over statistical and single machine learning-based models, achieving improved forecasting accuracy is still a challenging task. In the fuzzy time series model, the lengths of interval and fuzzy relationship group considered to be important factors that influence the forecasting accuracy of model. So, this research presents an FTS forecasting model based on Graph-Based Clustering technique (GBC). The graph-based clustering, which is an algorithm for data clustering, is untiled at the fuzzification stage to obtain unequal length of intervals. The proposed model is applied to forecast two numerical datasets as enrolments data of the University of Alabama and dataset of number of new confirmed cases of COVID-19 in Vietnam. The forecasting results obtained from the proposed model are compared to those produced by the other models at forecasting enrollments at the University of Alabama. It is observed that the proposed model achieves higher forecasting accuracy than its counterparts for all orders of fuzzy relationship.