Flood is one of the most devastating natural disasters that occurs frequently and can affect large community areas. It has become the main threat to people’s life and properties. This problem is more acute in highland areas of Ethiopia under strong environmental degradation due to population pressure. That is why, flood prediction has long been a popular subject matter to researchers around the world. Among other flood prediction alternatives, fuzzy logic is a formal attempt to capture, represent and work with objects with unclear or ambiguous boundaries. The main objective of the research is to design a flood predictive model that deal with reasoning that is approximate in the fuzzy environment. A methodology was proposed and applied to assesses the potential of fuzzy logic approach for real time flood level prediction using Mamdani fuzzy inference system. The main contribution of the research is to capture, both fuzziness and uncertainty inherent in the input/output data during prediction by apply Re-Partitioning Discretization(RPD) algorithm that, optimized the interval length between each fuzzy sets and significantly improves the prediction accuracy of the model. The study also has contributions in reduction of losses of life and property by alerting the community in the study area based on the outputs of the prediction model via SMS automatically to take immediate actions before the disaster occurs, assist local authorities, government practitioners, flood risk forecasting and warning authorities to accelerate the process of evacuating flood victims to the relief center. The accuracy of the prediction model, which is 91.57%, obtained after the validation of the model using test dataset showed a general agreement with the results from the measured value of the output. This also shows the acceptable levels of the prediction model, and which implies the potential of establishing a flood prediction model by using the fuzzy logic approach in the study area.