Data mining is a process of data grouping or partitioning from the large and complex data, and the clustering analysis is an important research field in data mining. The K-means algorithm is considered to be the most important unsupervised machine learning method in clustering, which can divide all the data into k subclasses that are very different from each other. By constantly iterating, the distance between each data object and the center of its subclass is minimized. Because K-means algorithm is simple and efficient, it is applied to data mining, knowledge discovery and other fields. However, the algorithm has its inherent shortcomings, such as the K value in the K-means algorithm needs to be given in advance; clustering results are highly dependent on the selection of initial... clustering centers and so on. In order to adapt to the historical data clustering of the geological disaster monitoring system, this paper presents a method to optimize the initial clustering center and the method of isolating points. The experimental results show that the improved k-means algorithm is better than the traditional clustering in terms of accuracy and stability, and the experimental results are closer to the actual data distribution.