The advancement of materials discovery through the applied artificial intelligence

In the past decade, several new informatics models have been introduced in materials science, offering rapid predictions instead of starting from scratch. Our research aims to achieve the optimization between targeted yield and likelihood. We gathered a matrix of 565 × 264 (264 features and 565 records) of laboratory data and conducted a series of experiments supported by various machine learning (ML) design patterns. Our research investigates the methods in the areas of 1). data foundation (data creation, data cleansing, data validation, and encoding/decoding); 2). model evaluations, including XGBoost (XGB), random forest (RF), and support vector machines (SVMs); 3). feature evaluation (clustering, feature selection, and feature weight evaluation); and 4). Bayesian optimization, optimization, and masking processes. Through this research, we explore the genuine methods in Materials Informatics (MI), and our methods such as Regression Deviation Degree Threshold (RDDT) and R:M ratio in data quality and measurement, and masking process in optimization process are novel in MI.