大規模住宅建設における炭素フットプリント最小化のためのグリーン建材選定への機械学習適用
Machine Learning for Green Building Material Selection to Minimize Carbon Footprint in Large Residential Construction (原題)
Aditya Ravindra Pawar, A. Zuting, Arjun Dhanjay Shedage, Gautam Shivaji Chindhe, Dipannita Mondal
🤖 gxceed AI 要約
日本語
建設分野のembodied carbon（建材由来排出）に着目し、建材の組合せを炭素影響カテゴリに分類する機械学習フレームワークを提案。ICEデータベースとインド・プネー地域の建設パラメータを統合した520シナリオでXGBoost、Random Forest、Decision Treeを比較し、XGBoostが精度94.2%・マクロF1 0.942で最高性能。SHAP分析で壁材・セメント種別・構造システムが最重要因子と判明。Power BIダッシュボードでリアルタイム予測と炭素削減比較を提供する。
English
Proposes an ML framework to classify construction material combinations into carbon-impact categories, targeting embodied carbon in large residential projects. Using 520 scenarios built from ICE carbon-intensity data and Pune, India construction parameters, XGBoost outperformed Random Forest and Decision Tree (94.2% accuracy, 0.942 macro F1). SHAP analysis identifies wall material, cement type, and structural system as key drivers, with a Power BI dashboard enabling real-time predictions and carbon-savings comparisons.
Unofficial AI-generated summary based on the public title and abstract. Not an official translation.
📝 gxceed 編集解説 — Why this matters
日本のGX文脈において
日本ではScope 3カテゴリ1（購入資材）や建設業の脱炭素開示が課題となる中、設計段階で建材由来排出を定量化する手法はSSBJ・有報のScope 3算定高度化に示唆を与える。地域パラメータ依存のため日本適用には国産建材DBとの接続が前提となる。
In the global GX context
Embodied carbon in construction is a growing focus under CSRD/ESRS E1 and Scope 3 Category 1 disclosure; ML-assisted material selection at design stage offers a replicable template for quantifying upstream emissions. The India-specific dataset limits direct transferability but demonstrates how interpretable ML (SHAP) can operationalize carbon accounting for built-environment disclosures.
👥 読者別の含意
🔬研究者:建材選択とembodied carbon分類にMLとSHAPを組み合わせた実証例として、AI×GX研究の設計参考になる。
🏢実務担当者:設計初期段階で建材組合せの炭素影響を比較・可視化するダッシュボード手法を自社調達・Scope 3算定に応用できる。
🏛政策担当者:建設分野のembodied carbon開示・算定標準化を検討する際、設計段階のML支援ツールの有用性と限界を示す。
📄 Abstract（原文）
The construction sector is widely recognized as one of the largest contributors to global greenhouse gas emissions, with a substantial portion arising from the production and use of construction materials rather than operational energy consumption alone [1], [5]. While advancements in building design have improved energy efficiency during the operational phase, emissions associated with material extraction, manufacturing, transportation, and construction—collectively referred to as embodied carbon—remain insufficiently addressed in early project planning. This study presents a machine learning–based framework aimed at supporting sustainable material selection for large residential construction projects. The proposed system evaluates combinations of construction materials and classifies them into predefined carbon impact categories, enabling decision-makers to understand environmental implications at the design stage. A dataset consisting of 520 building scenarios was developed by integrating carbon intensity values from the Inventory of Carbon and Energy (ICE) database with region-specific construction parameters relevant to Pune, India. Three supervised machine learning models—XGBoost, Random Forest, and Decision Tree—were implemented and evaluated. Among these, the XGBoost model achieved the highest performance, with an accuracy of 94.2% and a macro F1-score of 0.942, demonstrating its effectiveness in handling structured construction data. To enhance interpretability, SHAP (SHapley Additive exPlanations) analysis was employed to identify the relative importance of input features, revealing that wall material, cement type, and structural system are the most influential factors affecting carbon classification [13]. An interactive Power BI dashboard delivers real-time predictions, material comparisons, and carbon savings, helping construction professionals make informed, sustainable decisions while integrating machine learning with tools to reduce environmental impact.
🔗 Provenance — このレコードを発見したソース
- semanticscholar https://doi.org/10.65521/mjret.v13i2s.4062first seen 2026-10-21 05:25:29 · last seen 2026-11-06 05:37:46
🔔 こうした論文の新着を逃したくない方は キーワードアラート に登録（無料・3キーワードまで）。
gxceed は公開メタデータに基づく研究支援データセットです。要約・翻訳・解説は AI 支援で生成されています。 最終的な解釈・検証は利用者が原典資料に基づいて行うことを前提とします。