A rainstorm prediction scheme for Guangxi has been developed based on the European Centre for Medium-Range Weather Forecasting (ECMWF) products and using Particle Swarm Optimization-Neural Network (PNN) model. The PNN model input is constructed from potential predictors by employing both a stepwise regression method and an Isometric mapping (ISOMAP) algorithm of Manifold Learning. The ISOMAP algorithm is capable of finding meaningful low-dimensional architectures hidden in their nonlinear high-dimensional data space and separating the underlying factors. In this scheme, the new developed model, which is termed the PNN-ISOMAP model, is used for 24h rainstorm prediction in Guangxi. Using identical modeling samples and independent samples, predictions of five year spanning 2013-2017 based on the PNN-ISOMAP model are compared with the interpretation and application of ECMWF products in terms of the performance of rainstorm prediction at 89 stations in Guangxi. Results show that the PNN-ISOMAP model is superior to the interpolation method by ECMWF for rainstorm prediction, and the new scheme tends to have higher prediction accuracy, a stable prediction capability and robust generalization capability. Further, predictions of large-scale rainfall area show that the forecasting effect of the new scheme is more accurate than that of the interpolation method by ECMWF. The new scheme opens up a vast range of possibilities for operational weather prediction and merits further exploration.