One of the main problems that traditional support vector machine (SVM) has to solve is how to dynamically determine the kernel parameters and penalty parameters of the kernel function in time, along with the increasing amount of data and the changing data structure and characteristics. A new method is proposed for dynamic acquisition of SVM parameters by fruit fly optimization algorithm (FOA) based on the analysis of the classification and aggregation of land use data in urban industry. FOA-SVM aims at the relationship between feature words in the classification process and the core words of different activity semantics in context. In an incomplete date set of initial feature words, FOA-SVM can extract new feature words from the semantic association of feature words to improve the feature word date set. The dynamic parameters of SVM can be obtained through continuous training with FOA, and the accuracy of classification can be improved. The experimental results showed that FOA-SVM can process multi-feature synchronous classification according to different activity semantics and efficiently control the operation of the whole classification process, so as to obtain higher classification accuracy and stronger robustness in multi-source web date categorization. The efficiency of land use real-time analysis is improved.

传统支持向量机（SVM）必须解决的主要问题之一是如何随着数据量的增加以及数据结构和特性的变化而及时动态地确定内核函数的内核参数和惩罚参数。在分析城市工业用地数据分类与汇总的基础上，提出了一种利用果蝇优化算法动态获取支持向量机参数的新方法。FOA-SVM针对分类过程中的特征词与上下文中不同活动语义的核心词之间的关系。在初始特征词的日期不完整的情况下，FOA-SVM可以从特征词的语义关联中提取新的特征词，以改善特征词的日期集。通过FOA的连续训练可以得到SVM的动态参数，可以提高分类的准确性。实验结果表明，FOA-SVM可以根据不同的活动语义处理多特征同步分类，并有效控制整个分类过程的操作，从而在多源Web数据分类中获得更高的分类精度和较强的鲁棒性。提高了土地利用实时分析的效率。从而在多源Web日期分类中获得更高的分类精度和更强的鲁棒性。提高了土地利用实时分析的效率。从而在多源Web日期分类中获得更高的分类精度和更强的鲁棒性。提高了土地利用实时分析的效率。