Designing A Customized Search Engine For Precedent Analysis In The Architectural Design Process

MSTAS 2025 – The 19th Symposium on Digital Design in Architecture, Antalya, Turkey, June 30, 2025.

#ArchitecturalPrecedentAnalysis #ProjectCaseStudies #CustomizedSearchEngine #DataMining

The analysis of precedent projects in architectural design constitutes a critical phase for developing design knowledge through learning from existing examples. However, widely used digital architecture platforms today often fail to adequately support users in searching for projects based on specific criteria and accessing similar examples aligned with their research focus. Considering the limitations of these platforms in providing users with fast and accurate access to content matching their specific search criteria, this study addresses the design of a customized search engine tailored to user needs. The proposed search engine aims to enable filtering projects based on specific parameters and listing related examples. In this context, a data mining approach was adopted to uncover latent relationships between projects, employing the K-Means clustering algorithm, which offers rapid, scalable, and interpretable results.

The study is structured around three main phases: In the Data Collection and Organization (I) phase, information on architectural projects to be analysed was systematically collected and stored. A MySQL database was created to store the data. Project information was automatically retrieved from online architecture platforms using web scraping techniques developed in Python, and missing or unsuitable data for analysis was manually completed. The collected data was structured under 14 key attributes, including architect, function, location, topographical features, parcel type, building type, materials used, etc.. In the Data Analysis (II) phase, the aim was to process the collected data to reveal patterns and latent relationships between projects. Clustering analysis, a method of data mining, was applied for this purpose. Pre-processing techniques such as digitization and normalization were carried out to prepare the data for analysis. Subsequently, projects were clustered using the K-Means algorithm. The resulting high-dimensional data was reduced to two dimensions through dimensionality reduction and visualized via a scatter plot. In the Search Engine and User Interface Development (III) phase, a search engine and user interface were developed to present the analysed data to users in a meaningful and interactive manner. Application development and visual prototyping methods were employed in this process. The search engine, developed using Python and Flask, matches user queries with analysis results and lists related projects accordingly. The user interface was visualized using Figma and structured to form the basis for advanced modules such as filtering options and an AI-supported chatbot.

This system offers a solution to the limitations of current digital architecture platforms by establishing qualitative connections between projects and providing users with more meaningful and guided search results. It represents the first prototype of a personalized research tool that can be used in both educational and professional design environments. Although still in its prototypical phase, the system is open to further development through the expansion of the dataset, integration of visual analysis tools, automated data retrieval, and the use of artificial intelligence techniques such as natural language processing. In this regard, it has the potential to evolve not only as a project search tool but also as an intelligent decision support system capable of guiding the architectural design process.

Home page.

Search page.

The filter system.