<![CDATA[Accurate forecasting of tuberculosis (TB) cases is essential for effective public health planning, particularly in regions such as Sabah, Malaysia, where TB remains a significant and persistent health concern. This study aims to improve the accuracy of fuzzy time series models by refining the construction of Fuzzy Logical Relationship Groups using a revised heuristic framework. The proposed approach embeds domain-informed rules to dynamically adjust the formulation of fuzzy relationships. It was applied to monthly tuberculosis case data from 2012 to 2019 and evaluated against both the original fuzzy time series model and a heuristic-based variant. The revised heuristic model achieved the best forecasting accuracy, recording a Mean Squared Error of 1315.0160, a Root Mean Square Error of 36.2631, a Mean Absolute Error of 0.0566, and a Mean Absolute Percentage Error of 0.0138 percent. These consistently lower error values confirm the superiority of the revised model compared to the benchmarks. The study demonstrates that incorporating refined heuristic strategies enables fuzzy time series models to capture the dynamic nature of disease data more effectively. However, the analysis is limited to univariate data (monthly tuberculosis cases), and future work should consider multivariate and hybrid approaches. This research contributes to the understanding by demonstrating that revised heuristic knowledge significantly enhances the predictive capability of fuzzy time series models. The findings provide more reliable forecasts for tuberculosis trends and establish a foundation for broader applications in infectious disease forecasting and healthcare analytics.]]>
