| 摘要: |
| 针对当前居住建筑空间品质提升的
迫切需求，本文致力于突破传统“功能房”
的思维定式，提升空间划分精度与动态适应
性。基于微观使用者行为关联性视角，提出
一种融合设计结构矩阵（DSM）与模糊聚类
算法（FCM）的定量化设计方法，通过构建
居住行为细分数据库及四维空间关联指标模
型，运用聚类算法生成模块度最优的住宅功
能模块。研究结果表明，相较于传统功能分
区，该方法生成的住宅模块在模块度、类型
多样性、组合性与层级性方面均展现出显著
优势，有效验证了行为关联聚类策略的优越
性。本文为居住空间设计提供了从静态分区
转向动态聚类的科学方法，不仅推动空间组
织理论的创新，更为模块化建筑设计实践提
供可量化、可复用的技术支撑。 |
| 关键词: 行为空间 模块化 设计结构矩
阵 模糊聚类算法 住宅套内空间 |
| DOI：10.13791/j.cnki.hsfwest.20250708001 |
| 分类号: |
| 基金项目:国家自然科学青年基金项目（5210080700） ；深圳市医养建筑重点实验室（筹建启动）（ZDSYS20210623101534001） |
|
| Research on the spatial correlation of behavioral patterns in residential buildings based onFuzzy Clustering Algorithms: A case study of interior residential units |
|
ZENG Fanbo
|
| Abstract: |
| The persistent shortcomings of conventional residential design, rooted in the static
partitioning of space into predetermined “functional rooms”, underscore an urgent need for a
paradigm shift toward more adaptive and behaviorally responsive architectural solutions. This study
addresses this challenge by proposing a fundamental reorientation in spatial organization: from static,
function-based zoning to dynamic clustering derived from the correlations between micro-user
behaviors. The primary objective is to enhance both the precision of spatial division and the dynamic
adaptability of residential layouts through a novel, quantitative design methodology. To achieve this,
the research integrates the Design Structure Matrix (DSM) and the Fuzzy C-Means (FCM) clustering
algorithm, constructing a scientifically rigorous framework to translate behavioral patterns into
optimal spatial modules.Current research in residential spatial organization follows two predominant
paths, each with significant limitations. First, a component-based path, focuses on the standardization
of physical elements like furniture and building components. While it provides a clear, hierarchical
system for spatial decomposition and addresses “how to construct precisely”, it inherently treats space
as an assembly of static objects, failing to respond to the dynamics of user behavior and needs.
Consequently, it cannot answer “why this organization is optimal”. Second, a behavior-based path,
seeks to establish spatial modules by analyzing behavioral units and their spatiotemporal patterns.
However, this approach has largely remained qualitative or conceptual, constrained by limitations in
behavior data acquisition and the lack of a quantitative, computable model to effectively convert
behavioral insights into spatial design. This gap often forces designers to revert to empirical
“functional bubble diagrams” without genuine scientific generation capabilities. Fundamentally, the
prevailing paradigm remains one of “naming rooms for functions” rather than “organizing space for
behaviors”. To bridge this critical gap, this study develops and applies an integrated analytical model
combining DSM and FCM. The process begins with the deconstruction of traditional functional
rooms into a fine-grained spectrum of micro-behaviors. A comprehensive database of residential
behaviors is constructed, alongside a four-dimensional spatial correlation index model that quantifies
the relationships between these behavioral spaces. The DSM is employed to systematically map and
structure these correlations. Subsequently, the FCM clustering algorithm processes this quantitative
dataset, grouping the subdivided behavioral spaces into modules based on the strength of their
interconnections. The clustering is optimized to identify the scheme with the highest modularity,
ensuring that the resulting groupings exhibit strong internal cohesion and clear external boundaries.
Applied specifically to residential indoor space design, the method decomposes activities into 35
distinct micro-behavioral types. Through the computational clustering process, these are synthesized
into 15 optimized general behavioral space modules (e. g., modules integrating closely linked
behaviors that might traditionally be separated across a “kitchen”, “dining”, and “storage” room). A
comparative evaluation between these newly generated modules and traditional functional zoning
reveals the clear superiority of the behavior-correlation strategy. The new modules demonstrate
significant advantages across multiple quantitative metrics: enhanced modularity (indicating betterinternal coherence), greater typological diversity, improved combinability for flexible layout generation, and a more rational hierarchical structure. This
evidence effectively verifies that a clustering strategy driven by behavioral correlations yields a spatial organization system with inherently better design
properties than one based on a priori functional labels.The core contribution of this research is the proposal, development, and validation of the “behaviorcorrelation
clustering” method as a scientifically robust alternative to traditional functional zoning. By shifting the foundational unit of design from the room to
the behavior pattern, the study provides a quantifiable and reusable technical support system for modular architectural design practice. It advances spatial
organization theory by introducing a model that clarifies behavior types, quantifies spatial precision, and scientifically manages correlations. The constructed
hierarchical structure of behavioral space offers a practical reference framework for architectural design, one that can be applied to modular product design and
continuously refined with dynamic market or user data.Finally, this study establishes a generalizable model for behavioral space analysis. Future research will
focus on the empirical validation of design quality improvements through application in typical residential units, assessing outcomes in spatial efficiency,
behavior-flow fit, and user adaptability. Furthermore, the model’s flexibility allows for the integration of survey data from specific user groups as new
parameters, enabling highly customized spatial design and significantly expanding the methodology’s applicability. This pathway promises to transform
residential space design from a practice of static compartmentalization into a responsive, user-centered, and scientifically-informed process. |
| Key words: behavioral space modularity Design Structure Matrix (DSM) Fuzzy C-Means Clustering (FCM) interior residential unit space |
