Abstract: Digital building information modeling is existed in and applied to areas of building designing and urban planning. Due to the amount of data is increasing sharp today, it is necessary to classifying the information model based on content in order to overcome the phenomenon of the enrich data and poor knowledge. The paper proposes a method of model classification which based on classification of SVM decision model that combined the space syntax theory. It builds the Room Connectivity Attributed Relational Graphs（RCARG） for building information modeling, extracts the inherent characteristics of model Room and combines the model space configuration feature which enriched by the space syntax theory, adds the characteristic vector of the equalization process based on DAG - SVMS classification algorithm, reduces the misjudgment of decision-making risk classification to realize high precision accuracy of classification result. The experimental results show that the proposed method has higher classification accuracy compared with KNN and DAG-SVMS algorithm.

Key words: building information modeling, room connectivity attributed relational graphs, space syntax, DAG-SVMS

摘要： 数字化的建筑信息大量存在和应用于建筑设计、城市规划等领域。目前，由于建筑信息模型的数据量急剧膨胀，为克服“数据丰富而知识匮乏”现象，对其进行基于内容的模型分类十分必要。提出一种结合空间句法理论和基于SVM决策分类的模型分类方法，首先对建筑信息模型建立RCARG（Room Connectivity Attributed Relational Graphs）模型，提取出建筑信息模型的模型固有特征，并结合空间句法理论而扩充出模型空间构形特征，在常用的DAG-SVMS分类算法的基础上增加特征向量均衡化的过程，减少决策分类时误判几率，以实现高精准度分类效果。实验结果表明，该方法与KNN和DAG-SVMS算法相比，具有较高的分类精准度。

关键词: 建筑信息模型, 房间连接属性关系图（RCARG）, 空间句法, DAG-SVMS算法