BIM协同设计网络分析与情感识别研究

贺 领

摘 要

现如今，BIM技术的发展越来越迅速，BIM技术的应用也逐渐贯穿建筑的整个生命周期中。也正是因为BIM的应用的深入，人与模型、基于模型的人与人之间的交互也越来越多。但是这些交互的文本数据中隐藏的知识和经验没有被很好的挖掘，以至于这些用于辅助决策的优质素材白白浪费。

所以本文以期利用LDA模型和社会网络分析方法对基于BIM模型的协同设计中的通信文本进行知识挖掘，利用情感分析的方法对运维阶段公众的评论进行情感倾向判断，从而辅助项目管理者进行关键决策。具体如下：

本文利用Gephi软件构建了参与协同设计的所有用户交互的社会网络图谱，揭示了协同设计过程中相关人员的通信模式，根据凝聚子群分析和个人中心性分析发现了团队之间协同关系和核心人物，并对其中团队及个人的评论内容进行LDA主题挖掘分析，挖掘各个团队的关注焦点以及各相关人员利益关注点，从而达到了辅助决策者决策的目的。

本文采用了基于情感词典的情感分析方法对运维阶段业主和公众的建筑使用体验评论进行知识挖掘。首先利用中科院的ICTCLAS分词系统对文本进行预处理，并通过词性模板的方法，提取文本中的特征词和情感词，再通过对情感词的人工分类建立起了酒店领域的情感词典，而后根据其他研究构建了否定词集和带有权值的程度副词集，最后利用短语识别算法对各特征词以及属性类型进行了情感倾向打分，并将情感评价短语以可视化词云图的方式展示出来，为建筑工程运维阶段管理提供相关建议和决策支持。

关键词：BIM协同设计 知识挖掘  情感分析  社会网络分析   LDA模型  

Abstract

Nowadays, the development of BIM technology is getting faster and faster. And the application of BIM technology is gradually running through the entire life cycle of the building. So the interaction of people and models is getting more and more. However, the hidden knowledge and experience in these interactive text data is not well mined, so that the quality materials that should be used to aid decision-making are wasted.

Therefore, this thesis intends to use the LDA model and social network analysis method to mine the communication text in the collaborative design based on BIM, and use the sentiment analysis method to judge sentiment tendency of the public comments in the operation and maintenance stage, so as to assist the project manager to carry out the key decision-making. Details as follows:

This thesis uses Gephi software to construct a social network map of all user interactions involved in collaborative design, which reveals the communication mode of relevant personnel in the process of collaborative design and discovers the synergy between the teams and the core figures based on the analysis of cohesive subgroups and personal centrality. And this thesis carries out LDA topic modeling on the comments of the owner and public, which excavates the focus of each team and the interests of the relevant personnel. These results can assist decision-makers to optimize business processes and performance evaluation systems.

This thesis adopts the sentiment analysis method based on sentiment dictionary to discovery knowledge in the experience comments of the public during the operation and maintenance stage. Firstly, the text is preprocessed by the ICTCLAS word segmentation system of the Chinese Academy of Sciences. And the feature words and emotional words in the text are extracted by the POS template method. Then, the emotional dictionary of the hotel field is established by artificial classification of the emotional words. And according to other researches, a set of negative words and degree adverbs with weights are constructed. Finally, the phrase recognition algorithm is used to score the emotional words of each feature word and attribute type. And the emotional evaluation phrases are displayed in the form of visual word cloud map. The results provide relevant advice and decision-support for the managers during operation and maintenance stage of the building.

Keywords: Collaborative design based on BIM    Knowledge discovery  

Sentiment analysis   Social network analysis   LDA model