月別: 2021年8月

[Advanced Published online Journal of Computer Chemistry, Japan, by J-STAGE]
<Title:> First-Principles Calculations of Stability, Electronic Structure, and Sorption Properties of Nanoparticle Systems
<Author(s):> Gerardo VALADEZ HUERTA, Yusuke NANBA, Nor Diana Binti ZULKIFLI, David Samuel RIVERA ROCABADO, Takayoshi ISHIMOTO, Michihisa KOYAMA
<Corresponding author E-Mill:> koyama_michihisa(at)shinshu-u.ac.jp
<Abstract:> Nanoparticles have a wide range of applications as catalysts. Their catalytic and electronic properties differ from those of materials with flat surfaces and bulk materials. First-principles calculations of real system nanoparticles, which use nanoparticle models based on real shapes extracted from experimental observations, are essential for studying these properties to facilitate the computational design of new catalysts. In this article, we review first-principles studies of models of real systems of monometallic, bimetallic, and supported nanoparticles. The stability, electronic structure, hydrogen absorption behavior, and small molecule adsorption behavior are reviewed, and advances in first-principles calculations of real system nanoparticles are presented. Further, a combination of machine learning and first-principles studies is also considered. Future perspectives are discussed on the basis of these examples.
<Keywords:> Density functional theory, Real system structure, Nanoparticle, Machine learning
<URL:> https://www.jstage.jst.go.jp/article/jccj/advpub/0/advpub_2021-0028/_article/-char/ja/

[Advanced Published online Journal of Computer Chemistry, Japan, by J-STAGE]
<Title:> Constructing Regression Models with High Prediction Accuracy and Interpretability Based on Decision Tree and Random Forests
<Author(s):> Naoto SHIMIZU, Hiromasa KANEKO
<Corresponding author E-Mill:> hkaneko(at)meiji.ac.jp
<Abstract:> Models for predicting properties/activities of materials based on machine learning can lead to the discovery of new mechanisms underlying properties/activities of materials. However, methods for constructing models that exhibit both high prediction accuracy and interpretability remain a work in progress because the prediction accuracy and interpretability exhibit a trade-off relationship. In this study, we propose a new model-construction method that combines decision tree (DT) with random forests (RF); which we therefore call DT-RF. In DT-RF, the datasets to be analyzed are divided by a DT model, and RF models are constructed for each subdataset. This enables global interpretation of the data based on the DT model, while the RT models improve the prediction accuracy and enable local interpretations. Case studies were performed using three datasets, namely, those containing data on the boiling point of compounds, their water solubility, and the transition temperature of inorganic superconductors. We examined the proposed method in terms of its validity, prediction accuracy, and interpretability.
<Keywords:> Model interpretability, Predictive ability, Decision tree, Random forests, Regression model
<URL:> https://www.jstage.jst.go.jp/article/jccj/advpub/0/advpub_2020-0021/_article/-char/ja/

[Published online Journal of Computer Chemistry, Japan Vol.20, 14-21, by J-STAGE]
<Title:> 熱硬化性樹脂コンポジットにおける物性予測に向けた機械学習モデル構築
<Author(s):> 高原　渉, 小林　優希, 森田　将司, 奥山　浩二郎, 川村　信行
<Corresponding author E-Mill:> takahara.wataru(at)jp.panasonic.com
<Abstract:> 本研究では自社の実験データを用いて，熱硬化性樹脂コンポジットを工業応用する際に重要となる比誘電率(ε)，誘電正接(tanδ)予測に向けた機械学習モデルを構築した．機械学習モデルの構築には近年注目を集めている勾配ブースティング木(GBDT)系のアルゴリズムを含む幅広い手法を採用した．複数の手法にて構築したモデルの中で，Training data setにおける交差検証(Cross-validation)時の決定係数R²_CV > 0.8を満たすモデルを抽出した．更にTraining data set においてRMSE (Root Mean Square Error)及びMAE (Mean Absolute Error)の値が小さく，より定量的な物性予測が可能と考えられるモデルを選択し，Test data setにおける評価を行った．その結果，RMSEやMAEがε及びtanδそれぞれの平均値に対して10^-1 10^-2オーダーで物性予測可能な機械学習モデルが得られた．本結果より，熱硬化性樹脂コンポジットにおいてもMI (Materials Informatics)によるアプローチが有効であり，定量的な特性予測が可能であることを初めて実証した．今後の開発において，本アプローチを用いることで材料開発期間の短縮及び材料開発の促進を期待する．
<Keywords:>
<URL:> https://www.jstage.jst.go.jp/article/jccj/20/1/20_2021-0026/_article/-char/ja/