This paper proposes a novel vehicle speed prediction method based on a fuzzy Markov chain model. To enhance prediction accuracy, a triangular membership function is employed to map the acceleration into fuzzy states, enabling smooth transitions between adjacent states. The proposed method is evaluated under three driving cycles, demonstrating superior performance with average prediction error reductions of 4.28% and 3.41% compared to conventional multi-step Markov chains and long short-term memory neural networks, respectively. Furthermore, a comparative analysis with the Gaussian membership function confirms the effectiveness of the selected triangular function for this application. The results highlight the potential of fuzzy Markov chains in improving the precision of vehicle speed prediction.