On analyzing math learning and reasoning usinga data mining approach

Emilio Gerardo Sotto Riveros¹; Christian Schaerer²; Santiago Gomez Guerrero¹

¹ Polytechnic School, National University of Asuncion; ² Centro de Investigación en Matemática - CIMA, Polytechnic School, National University of Asuncion

doi:10.20906/CPS/CILAMCE2017-0655

Resumo

Learning process in mathematics is paramount for all science and engineering. It is fairly intuitive that several factors influence learning and reasoning, and specific learning issues may impact a person's professional life. Several works attribute the academic success of students to their teachers' prowess. This is important, but it has been empirically observed that under equal conditions, some students demonstrate better skills for speed learning. The reason for differences in math learning and reasoning among students is still not well understood. In this work we explore the use of data mining techniques for understanding aspects that most influence the process of mathematical learning. Math performance data are taken from the answers given to a battery of problems by students in the age range from 11 to 17 years old, grouped into scores that correspond to math subjects like algebra, geometry, statistics and so on. Each score is considered a response variable. We attempt to explain scores with a second dataset consisting of the same students' answers to questions in the areas of school life, study habits, perceptions related to teachers, socio-economic factors and the family environment, looking for relationships between the personal characteristics of the students and the quality of their reasoning in math. This is still an ongoing work, and preliminary results show that some scores get more influenced than others by variables in specific areas like, say, family environment. By contrast, some of these specific variables may seem unimportant for explaining total score.

Palavras-chave: analyzing; math; data maning; processing