Accuracy prediction using data-driven algorithm for carbon containing compounds. (2022)

• Record Type: Journal Article
• Authors: Bisht, Richa; Kumar, Deepak; Paswan, Manikant
• Is Part Of: Materials today. Volume 68:Part 6(2022)
• Page Start: 1921
• Page End: 1925
• Publication Date: 2022
• Subjects: Data-Driven -- Regression Model -- Carbon Compounds -- Materials; Materials science -- Congresses -- Periodicals; 620.1
• DOI: 10.1016/j.matpr.2022.08.119
• Languages: English
• ISSNs: 2214-7853

Abstract

Abstract: Technology is usually a solution to a problem and seeks materials with specific properties; therefore, materials are tailored to specific needs and serve a particular purpose. Therefore, careful research is needed to meet this growing demand for new materials. However, the conventional method of discovering materials has two key drawbacks: high cost and a lot of time. Materials research focuses on a more efficient side to address these obstacles. It is currently looking for data-driven, highly computational approaches with maximum accuracy to develop tailored materials in a short time. In recent years, work in this area has increased significantly, and some have made significant discoveries. Machine learning is an entirely data-driven technique that meets all these requirements equally. Various machine learning algorithms may be taught, and significant findings are produced from many computational and experimental data in materials science. In this work, datasets of carbonaceous compounds are used to predict the accuracy of data-driven machine learning algorithms. First, the optimal number of clusters is calculated using the elbow method. Then, the whole datasets are divided into four clusters. Afterward, the accuracy is predicted using the decision tree and random forest regression models for performance evaluation. It is found that the decision tree model achieves 99.4% accuracy compared to the random forest model (99.23%).