Scholarlink

This article introduces a data-driven methodology for the conceptual design of general education institutions using machine learning and generative modelling. The study evaluates the applicability of two computational approaches—linear regression and support vector method—implemented within the Grasshopper parametric environment for Rhinoceros. A total of 15 design iterations were tested using three variable parameters: link ratio (3:2:3/2:2:1/0:3:2), building height (1–4 floors), and maximum block width (11–19 m). Performance metrics included maximum wind load (3.84–4.59 m/s), average insolation (3.61–3.86 h), and algorithm execution time (8.28–9.15 min). The methodology was validated using a real construction site in St. Petersburg, where calculations confirmed the feasibility of a school building exceeding 10,800 m2. The results demonstrate that linear regression ensures more stable generation than the support vector method, enabling faster configuration formation and reducing manual design workload. The proposed framework offers a scalable foundation for AI-assisted architectural planning and supports future integration into BIM-oriented workflows and automated design systems.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

The Utilization of Machine Learning Algorithms in the Design of Educational Institutions

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Liliia Talipova,
  • Evangelina Morozova

摘要

This article introduces a data-driven methodology for the conceptual design of general education institutions using machine learning and generative modelling. The study evaluates the applicability of two computational approaches—linear regression and support vector method—implemented within the Grasshopper parametric environment for Rhinoceros. A total of 15 design iterations were tested using three variable parameters: link ratio (3:2:3/2:2:1/0:3:2), building height (1–4 floors), and maximum block width (11–19 m). Performance metrics included maximum wind load (3.84–4.59 m/s), average insolation (3.61–3.86 h), and algorithm execution time (8.28–9.15 min). The methodology was validated using a real construction site in St. Petersburg, where calculations confirmed the feasibility of a school building exceeding 10,800 m2. The results demonstrate that linear regression ensures more stable generation than the support vector method, enabling faster configuration formation and reducing manual design workload. The proposed framework offers a scalable foundation for AI-assisted architectural planning and supports future integration into BIM-oriented workflows and automated design systems.