ISODATA算法是一种基于统计模式识别的，非常经典的非监督学习动态聚类算法，有较强的实用性。ISODATA算法不仅可以通过调整样本所属类别完成样本的聚类分析，而且可以自动地进行类别的“合并”和“分裂”，从而得到类数比较合理的聚类结果。-ISODATA algorithm is based on statistical pattern recognition, unsupervised learning is the classic dynamic clustering algorithm, there is a strong practical. ISODATA algorithm can be completed by adjusting the sample of category clustering analysis of samples, and the categories can be automatically " merge"　and " split"　to get class numbers more reasonable clustering results.