From Data to Design:A Machine Learning-Based Decision Support System for LEED Certification Strategies

Cekmegeli S.

44th Conference on Education and Research in Computer Aided Architectural Design in Europe, eCAADe 2026, Lübeck, Germany, 7 - 11 September 2026, vol.1, pp.419-428, (Full Text)

  • Publication Type: Conference Paper / Full Text
  • Volume: 1
  • Doi Number: 10.52842/conf.ecaade.2026.1.419
  • City: Lübeck
  • Country: Germany
  • Page Numbers: pp.419-428
  • Keywords: decision support systems, early-stage design, LEED certification, machine learning, sustainable architecture
  • Middle East Technical University Affiliated: Yes

Abstract

Sustainability certification systems like LEED often function as retrospective compliance frameworks rather than proactive early-design guides. This research proposes a machine learning-based Decision Support System (ML-DSS) translating certification data into early-stage design guidance. The framework comprises a Random Forest classifier ranking LEED credits by discriminative importance, a K-Nearest Neighbors engine identifying similar target-level reference projects, and a proof-of-concept Knowledge Base translating credit gaps into spatial, technical, and procedural recommendations. The framework is version-agnostic, drawing on version- and typology-specific datasets of 19,163 certified BD+C projects filtered from 215,071 entries across LEED v3, v4, and v4.1. Applicability is demonstrated through two scenarios: a retrospective analysis of the Kalyon Karapınar SCADA Building, identifying three credit gaps relevant to its one-point Silver-to-Gold gap under LEED v4 BD+C: NC; and a Day Zero scenario for a hypothetical LEED 2009 Schools project targeting Gold, generating a pathway from early site, policy, and design inputs. Optimize Energy Performance ranked highest in both Random Forest models and was achieved by all five nearest Gold-certified schools despite being absent from that profile. These findings suggest certification opportunities can be identified before detailed design decisions are finalized, reframing certification as an early design resource rather than a later evaluation mechanism.