USING A MACHINE LEARNING APPROACH TO EXPLORE NON-COGNITIVE FACTORS AFFECTING READING, MATHEMATICS, AND SCIENCE LITERACY IN CHINA AND THE UNITED STATES

Abstract

Non-cognitive factors are considered critical aspects in shaping students' academic achievement. This study aims to analyze and explore the mechanisms of the influence of noncognitive factors on 15-year-old students' abilities in China and the United States. Based on the Programme for International Student Assessment (PISA) 2018 education dataset, the Classification and Regression Tree (CART) model identifies and explains the factors. The study finds that there are 11 most influential common features in China and 9 in the United States. The two countries have 5 common features, the meta-cognition assess credibility, summarizing text ability, PISA test difficulty perception, science learning time, and school lessons numbers per week. Family economic status also impacts personal ability. Regarding subject characteristics, attitude towards failure is the determinant of reading and mathematics. Cooperation and competition among students help to improve mathematics and science. Furthermore, the comparison between the two countries concludes that selfawareness, family economic status, and school learning environment are critical to personal ability. The study concludes that it is necessary to foster a sense of healthy competition among students at the school level and provide more attention to students with low family socioeconomic status to improve their abilities.