{"id":{"repo_id":"cambridge","oai_identifier":"oai:www.repository.cam.ac.uk:1810/387533"},"canonical_url":"https://search.dev.ndltd.org/etd/cambridge/oai:www.repository.cam.ac.uk:1810/387533","repository":{"repo_id":"cambridge","name":"Cambridge University","base_url":"https://api.repository.cam.ac.uk/server/oai/request"},"display":{"title":"A Systematic Framework for Feature Analysis, Selection, and Property Prediction of Chemical Materials Using Machine Learning","abstract":"This thesis delineates the development and application of a gradient-boosted statistical feature selection workflow within the field of materials informatics. Its utility is demonstrated via the prediction of various material properties across different research domains, including solid-state physics, condensed matter physics, materials science and engineering, as well as spectroscopic characterisation. The overarching objective is to establish a holistic, universally applicable materials discovery workflow that minimises computational overhead, while mitigating potential human biases throughout the model development stage. Such a general-purpose workflow is adept at accommodating both organic and inorganic materials, thereby establishing it as a versatile and adaptable tool for diverse material inputs. Chapter 1 reviews the progress in the materials discovery process and the pivotal role that machine learning has played in these advancements. It examines the shortcomings that are associated with current prediction methods in materials informatics. This discussion sets the stage for the development of a systematic approach to feature analysis, selection, and modelling of chemical materials aimed at enhancing materials discovery. Chapter 2 delves into the detailed components of the proposed workflow, demonstrating its effectiveness in predicting material properties. The underlying theoretical concepts and essential background information that support the development and application of the workflow are presented therein. In Chapter 3 and \\Chapter 4, the application of the proposed workflow to predict the properties of inorganic materials is presented. A comprehensive study is conducted on predicting the Curie temperature based on the chemical composition of ferromagnetic materials. Additionally, the workflow demonstrates its capability to autonomously construct magnetic phase diagrams by extrapolating relationships learned from a curated materials database. The application of the workflow is further extended across diverse research domains, showcasing its ability to accurately predict material properties when benchmarked against high-throughput calculations and experimental measurements. This is explored in subsequent subsections, encompassing the prediction of band gaps, predictive modelling of high-entropy alloys and amorphous metallic alloys, as well as the estimation of critical temperatures for superconductors. In Chapter 5, the focus is shifted to the prediction of material properties in organic materials. Building on the details of the workflow presented previously, this chapter highlights its predictive capabilities across a broad range of organic properties, including drug-related attributes and quantum mechanical characteristics. A key demonstration is the integration of the workflow with other machine learning algorithms, such as unsupervised learning methods that generate substructure vector embeddings. Additionally, a study is presented where convolutional neural networks are used in conjunction to predict optical absorption peaks in organic materials, showcasing the workflow's compatibility with advanced machine learning techniques. Lastly, Chapter 6 concludes this thesis and suggests future research topics that could complement or further broaden the scope of predictive modelling within the field of materials informatics.","abstract_html":"This thesis delineates the development and application of a gradient-boosted statistical feature selection workflow within the field of materials informatics. Its utility is demonstrated via the prediction of various material properties across different research domains, including solid-state physics, condensed matter physics, materials science and engineering, as well as spectroscopic characterisation. The overarching objective is to establish a holistic, universally applicable materials discovery workflow that minimises computational overhead, while mitigating potential human biases throughout the model development stage. Such a general-purpose workflow is adept at accommodating both organic and inorganic materials, thereby establishing it as a versatile and adaptable tool for diverse material inputs. Chapter 1 reviews the progress in the materials discovery process and the pivotal role that machine learning has played in these advancements. It examines the shortcomings that are associated with current prediction methods in materials informatics. This discussion sets the stage for the development of a systematic approach to feature analysis, selection, and modelling of chemical materials aimed at enhancing materials discovery. Chapter 2 delves into the detailed components of the proposed workflow, demonstrating its effectiveness in predicting material properties. The underlying theoretical concepts and essential background information that support the development and application of the workflow are presented therein. In Chapter 3 and \\Chapter 4, the application of the proposed workflow to predict the properties of inorganic materials is presented. A comprehensive study is conducted on predicting the Curie temperature based on the chemical composition of ferromagnetic materials. Additionally, the workflow demonstrates its capability to autonomously construct magnetic phase diagrams by extrapolating relationships learned from a curated materials database. The application of the workflow is further extended across diverse research domains, showcasing its ability to accurately predict material properties when benchmarked against high-throughput calculations and experimental measurements. This is explored in subsequent subsections, encompassing the prediction of band gaps, predictive modelling of high-entropy alloys and amorphous metallic alloys, as well as the estimation of critical temperatures for superconductors. In Chapter 5, the focus is shifted to the prediction of material properties in organic materials. Building on the details of the workflow presented previously, this chapter highlights its predictive capabilities across a broad range of organic properties, including drug-related attributes and quantum mechanical characteristics. A key demonstration is the integration of the workflow with other machine learning algorithms, such as unsupervised learning methods that generate substructure vector embeddings. Additionally, a study is presented where convolutional neural networks are used in conjunction to predict optical absorption peaks in organic materials, showcasing the workflow's compatibility with advanced machine learning techniques. Lastly, Chapter 6 concludes this thesis and suggests future research topics that could complement or further broaden the scope of predictive modelling within the field of materials informatics.","abstract_has_math":false,"creators":["Jung, Son Gyo"],"institution":"University of Cambridge","degree_name":"Doctor of Philosophy (PhD)","degree_level":"Doctoral","degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":["Cole, Jacqueline"],"committee_chairs":[],"committee_members":[],"year":2025,"date_issued":"2025-02-28","date_published":"2025-02-28","updated_at":"2026-07-22T22:24:24Z","subjects":["chemistry","machine learning","materials informatics","physics"],"languages":[],"rights":[],"rights_urls":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/5f5ae7ce-c4c2-49d6-9893-e0535ddf57ab/download","http://purl.org/NET/rdflicense/allrightsreserved"],"identifier_entries":[]},"links":{"outbound_url":"https://doi.org/10.17863/CAM.120278","outbound_label":"DOI","outbound_source":"dc:identifier.doi"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Cole, Jacqueline"]},{"key":"dc:contributor.sponsor","label":"Sponsor","values":["Science and Technology Facilities Council (STFC) via the ISIS Neutron and Muon Source; University of Cambridge"]},{"key":"dc:creator","label":"Author","values":["Jung, Son Gyo"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.issued","label":"Date","values":["2025-02-28"]},{"key":"dc:publisher.institution","label":"Dc Publisher Institution","values":["University of Cambridge"]},{"key":"dc:relation.isreferencedby.uri","label":"Dc Relation Isreferencedby URI","values":["https://www.repository.cam.ac.uk/handle/1810/387533"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]},{"key":"dc:type.qualificationlevel","label":"Dc Type Qualificationlevel","values":["Doctoral"]},{"key":"dc:type.qualificationname","label":"Dc Type Qualificationname","values":["Doctor of Philosophy (PhD)"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["chemistry","machine learning","materials informatics","physics"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:rights","label":"Dc Rights","values":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/5f5ae7ce-c4c2-49d6-9893-e0535ddf57ab/download","http://purl.org/NET/rdflicense/allrightsreserved"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.doi","label":"DOI","values":["https://doi.org/10.17863/CAM.120278"]},{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/e390ef14-4101-492f-8930-492b526cd88b/download"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["This thesis delineates the development and application of a gradient-boosted statistical feature selection workflow within the field of materials informatics. Its utility is demonstrated via the prediction of various material properties across different research domains, including solid-state physics, condensed matter physics, materials science and engineering, as well as spectroscopic characterisation. The overarching objective is to establish a holistic, universally applicable materials discovery workflow that minimises computational overhead, while mitigating potential human biases throughout the model development stage. Such a general-purpose workflow is adept at accommodating both organic and inorganic materials, thereby establishing it as a versatile and adaptable tool for diverse material inputs. Chapter 1 reviews the progress in the materials discovery process and the pivotal role that machine learning has played in these advancements. It examines the shortcomings that are associated with current prediction methods in materials informatics. This discussion sets the stage for the development of a systematic approach to feature analysis, selection, and modelling of chemical materials aimed at enhancing materials discovery. Chapter 2 delves into the detailed components of the proposed workflow, demonstrating its effectiveness in predicting material properties. The underlying theoretical concepts and essential background information that support the development and application of the workflow are presented therein. In Chapter 3 and \\Chapter 4, the application of the proposed workflow to predict the properties of inorganic materials is presented. A comprehensive study is conducted on predicting the Curie temperature based on the chemical composition of ferromagnetic materials. Additionally, the workflow demonstrates its capability to autonomously construct magnetic phase diagrams by extrapolating relationships learned from a curated materials database. The application of the workflow is further extended across diverse research domains, showcasing its ability to accurately predict material properties when benchmarked against high-throughput calculations and experimental measurements. This is explored in subsequent subsections, encompassing the prediction of band gaps, predictive modelling of high-entropy alloys and amorphous metallic alloys, as well as the estimation of critical temperatures for superconductors. In Chapter 5, the focus is shifted to the prediction of material properties in organic materials. Building on the details of the workflow presented previously, this chapter highlights its predictive capabilities across a broad range of organic properties, including drug-related attributes and quantum mechanical characteristics. A key demonstration is the integration of the workflow with other machine learning algorithms, such as unsupervised learning methods that generate substructure vector embeddings. Additionally, a study is presented where convolutional neural networks are used in conjunction to predict optical absorption peaks in organic materials, showcasing the workflow's compatibility with advanced machine learning techniques. Lastly, Chapter 6 concludes this thesis and suggests future research topics that could complement or further broaden the scope of predictive modelling within the field of materials informatics."]},{"key":"dc:format.checksum.md5","label":"Dc Format Checksum Md5","values":["784e8bfacef6eabfe4057d939bec9e68","87eda9de84448d1f82354d60eee3eb5f"]},{"key":"dc:title","label":"Title","values":["A Systematic Framework for Feature Analysis, Selection, and Property Prediction of Chemical Materials Using Machine Learning"]}]}],"canonical_facts":{"dc:contributor.advisor":["Cole, Jacqueline"],"dc:contributor.sponsor":["Science and Technology Facilities Council (STFC) via the ISIS Neutron and Muon Source; University of Cambridge"],"dc:creator":["Jung, Son Gyo"],"dc:date.issued":["2025-02-28"],"dc:description.abstract":["This thesis delineates the development and application of a gradient-boosted statistical feature selection workflow within the field of materials informatics. Its utility is demonstrated via the prediction of various material properties across different research domains, including solid-state physics, condensed matter physics, materials science and engineering, as well as spectroscopic characterisation. The overarching objective is to establish a holistic, universally applicable materials discovery workflow that minimises computational overhead, while mitigating potential human biases throughout the model development stage. Such a general-purpose workflow is adept at accommodating both organic and inorganic materials, thereby establishing it as a versatile and adaptable tool for diverse material inputs. Chapter 1 reviews the progress in the materials discovery process and the pivotal role that machine learning has played in these advancements. It examines the shortcomings that are associated with current prediction methods in materials informatics. This discussion sets the stage for the development of a systematic approach to feature analysis, selection, and modelling of chemical materials aimed at enhancing materials discovery. Chapter 2 delves into the detailed components of the proposed workflow, demonstrating its effectiveness in predicting material properties. The underlying theoretical concepts and essential background information that support the development and application of the workflow are presented therein. In Chapter 3 and \\Chapter 4, the application of the proposed workflow to predict the properties of inorganic materials is presented. A comprehensive study is conducted on predicting the Curie temperature based on the chemical composition of ferromagnetic materials. Additionally, the workflow demonstrates its capability to autonomously construct magnetic phase diagrams by extrapolating relationships learned from a curated materials database. The application of the workflow is further extended across diverse research domains, showcasing its ability to accurately predict material properties when benchmarked against high-throughput calculations and experimental measurements. This is explored in subsequent subsections, encompassing the prediction of band gaps, predictive modelling of high-entropy alloys and amorphous metallic alloys, as well as the estimation of critical temperatures for superconductors. In Chapter 5, the focus is shifted to the prediction of material properties in organic materials. Building on the details of the workflow presented previously, this chapter highlights its predictive capabilities across a broad range of organic properties, including drug-related attributes and quantum mechanical characteristics. A key demonstration is the integration of the workflow with other machine learning algorithms, such as unsupervised learning methods that generate substructure vector embeddings. Additionally, a study is presented where convolutional neural networks are used in conjunction to predict optical absorption peaks in organic materials, showcasing the workflow's compatibility with advanced machine learning techniques. Lastly, Chapter 6 concludes this thesis and suggests future research topics that could complement or further broaden the scope of predictive modelling within the field of materials informatics."],"dc:format.checksum.md5":["784e8bfacef6eabfe4057d939bec9e68","87eda9de84448d1f82354d60eee3eb5f"],"dc:identifier.doi":["https://doi.org/10.17863/CAM.120278"],"dc:identifier.uri":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/e390ef14-4101-492f-8930-492b526cd88b/download"],"dc:publisher.institution":["University of Cambridge"],"dc:relation.isreferencedby.uri":["https://www.repository.cam.ac.uk/handle/1810/387533"],"dc:rights":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/5f5ae7ce-c4c2-49d6-9893-e0535ddf57ab/download","http://purl.org/NET/rdflicense/allrightsreserved"],"dc:subject":["chemistry","machine learning","materials informatics","physics"],"dc:title":["A Systematic Framework for Feature Analysis, Selection, and Property Prediction of Chemical Materials Using Machine Learning"],"dc:type":["Thesis"],"dc:type.qualificationlevel":["Doctoral"],"dc:type.qualificationname":["Doctor of Philosophy (PhD)"]},"updated_at":"2026-07-22T22:24:24Z"}