The establishment of natural disaster prediction model to predict and analyze the occurrence of natural disasters is conducive to enhance the technical level of disaster prevention and mitigation. We present a design and implementation of natural disaster forecasting system based on association rules and the Web text mining technology. The system uses a mature open source crawler framework to capture the unstructured natural disaster information from the authoritative disaster information release platforms. By using the Chinese word segmentation technique the data are cleaned up and organized into a structured natural disaster database, and mining association rules of natural disaster events are worked out from improved association rule algorithm. And then by monitoring the front-end information of association rules in real time, the prediction of natural disasters can be achieved. Experimental results show that the system can effectively mine the association rules of natural disaster information with high confidence.