In this paper, a new heuristic learning algorithm, the modified water flow-like algorithm (MWFA), is proposed for TSK-type interval-valued fuzzy system (TIVFS) optimization. The water flow-like algorithm (WFA) is inspired by the natural behavior of water flows; the splitting, moving, merging, evaporation, and precipitation operations were developed for optimization. However, the original WFA is not valid for the continuous optimization problem. Therefore, some modifications, including some moving strategies, such as applying the tabu searching and gradient-descent techniques, are proposed to enhance the performance of MWFA. In addition, the modified strategies in evaporation and precipitation operations are more consistent with the natural behavior of water flows and they increase the diversity of solution agents. Finally, the MWFA is applied in the design of TIVFS for nonlinear system identification to demonstrate the effectiveness and performance.