Aiming at the difficulty of hydrological modeling in hydrological data-scarce catchments, a regional flood forecasting method based on long-short term memory neural network (LSTM) was proposed in this paper. Through scalarizing the data of topographic and geomorphological factors of each catchment in the same hydroclimatic similarity zone, the influence of local factors could be eliminated. On this basis, a unified modeling dataset in a similarity zone was constructed and the sample size was effectively expanded, which provided a possibility for the establishment of flood forecasting models in the data-sparse catchments. The Jiaodong Peninsula was selected as the study area in this study. In order to verify the application effect of the regional model in different scenarios, two regional modeling schemes were designed in this paper. To be specific, the first scheme was to construct a regional model based on the data of other basins in the similarity zone, without the participation of the data of the forecast basin (regional model Ⅰ). The second scheme was to construct a regional model with the data both from the forecast basin and other basins in the same region (regional model Ⅱ). In addition, the “single-basin model”, which was trained only with forecast basin's data, was selected as a benchmark model. Results indicated that, for data-scarce basins in this study, both regional models exhibited high accuracy and were superior to single-basin model. Comparatively, the regional model that incorporated data from the forecast basin outperformed the other regional model, suggesting that the accuracy of the regional model could be further improved if the data of the forecast basin was incorporated into model construction. The study can provide a reference for flood forecasting in data-scarce catchments.