Explainable Machine Learning Algorithms For Predicting Glass Transition Temperatures

Author(s)

Alcobaca, Edesio ^[1] ; Mastelini, Saulo Martiello ^[1] ; Botari, Tiago ^[1] ; Pimentel, Bruno Almeida ^[1] ; Cassar, Daniel Roberto ^[2] ; de Leon Ferreira de Carvalho, Andre Carlos Ponce ^[1] ; Zanotto, Edgar Dutra ^[2]

Affiliation

^[1] Univ Sao Paulo, Inst Math & Comp Sci, Sao Carlos - Brazil

^[2] Univ Fed Sao Carlos, Ctr Res Technol & Educ Vitreous Mat, Dept Mat Engn, Sao Carlos - Brazil

• Document type: Journal article
• Source: ACTA MATERIALIA; v. 188, p. 92-100, APR 15 2020.
• Web of Science Citations: 0

Abstract

Modern technologies demand the development of new glasses with unusual properties. Most of the previous developments occurred by slow, expensive trial-and-error approaches, which have produced a considerable amount of data over the past 100 years. By finding patterns in such types of data, Machine Learning (ML) algorithms can extract useful knowledge, providing important insights into composition-property maps. A key step in glass composition design is to identify their physical-chemical properties, such as the glass transition temperature, T-g. In this paper, we investigate how different ML algorithms can be used to predict the T-g of glasses based on their chemical composition. For such, we used a dataset of 43,240 oxide glass compositions, each one with its assigned T-g. Besides, to assess the predictive performance obtained by ML algorithms, we investigated the possible gains by tuning the hyperparameters of these algorithms. The results show that the best ML algorithm for predicting T-g is the Random Forest (RF). One of the main challenges in this task is the prediction of extreme T-g values. To do this, we assessed the predictive performance of the investigated ML algorithms in three T-g intervals. For extreme T-g values (<= 450 K and >= 1150 K), the top-performing algorithm was the k-Nearest Neighbours, closely followed by RF. The induced RF model predicted extreme values of T-g with a Relative Deviation (RD) of 3.5% for glasses with high T-g (>= 1150 K), and RD of 7.5% for glasses with very low T-g (<= 450 K). Finally, we propose a new visual approach to explain what our RF model learned, highlighting the importance of each chemical element to obtain glasses with extreme T-g. This study can be easily expanded to predict other composition-property combinations and can advantageously replace empirical approaches for developing novel glasses with relevant properties and applications. (C) 2020 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. (AU)

• FAPESP's process: 17/12491-0 - Kinetic processes in glass and formulation of new glasses using machine learning
• Grantee: Daniel Roberto Cassar
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• FAPESP's process: 13/07375-0 - CeMEAI - Center for Mathematical Sciences Applied to Industry
• Grantee: Francisco Louzada Neto
• Support Opportunities: Research Grants - Research, Innovation and Dissemination Centers - RIDC
• FAPESP's process: 18/07319-6 - Multi-target data stream mining
• Grantee: Saulo Martiello Mastelini
• Support Opportunities: Scholarships in Brazil - Doctorate
• FAPESP's process: 17/06161-7 - Interpretability of deep networks
• Grantee: Tiago Botari
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• FAPESP's process: 17/20265-0 - Use of meta-learning for clustering algorithm selection problems
• Grantee: Bruno Almeida Pimentel
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• FAPESP's process: 13/07793-6 - CEPIV - Center for Teaching, Research and Innovation in Glass
• Grantee: Edgar Dutra Zanotto
• Support Opportunities: Research Grants - Research, Innovation and Dissemination Centers - RIDC
• FAPESP's process: 18/14819-5 - Automated machine learning: learning to learn
• Grantee: Edesio Pinto de Souza Alcobaça Neto
• Support Opportunities: Scholarships in Brazil - Doctorate (Direct)