A female teacher wearing a face mask visits her classroom while learners are at home taking online classes or completing modules to continue their education amid the Covid-19 pandemic. In the Philippines’ first year of participating in the International Student Assessment Program, results showed that 15-year-old Filipino students ranked last in reading out of 77 countries that participated in the assessment. PHOTO BY RUY MARTINEZ
Data scientists at De La Salle University (DLSU) used machine learning (ML) to analyze factors that affect student performance using data on reading scores from the International Assessment Program of students (PISA) in 2018.
In the Philippines’ first year of participation in PISA, results showed that 15-year-old Filipino students ranked last in reading out of 77 countries that participated in the assessment.
The best models created by ML by the DLSU research team show 81% accuracy in determining whether a student profile can meet or may fall below the standard.
“We wanted to determine what are the factors that predict [which students will fail to meet the standards], and so we can identify who is vulnerable. And I think that if we know the common characteristics that characterize this group, we will have a better idea of the interventions needed, ”explained Professor Allan Bernardo, member of the research project and academic researcher in the Department of Psychology at DLSU.
The preliminary results of the study suggest factors in classifying students who meet proficiency standards versus those who do not: 1) students’ personal beliefs about themselves and about school, such as believing that they can no longer improve their reading skills and view school as an unimportant aspect of life; 2) whether or not they have a certain sense of belonging to the school or if they are victims of bullying; and 3) financial factors such as being of lower socio-economic status, lack of educational resources in schools, and having low career aspirations after graduation.
The team plans to discuss the findings with education stakeholders and policy makers, including the Department of Education (DepEd) and Congress.
“We are preparing the guidance note, and this will certainly include our recommendations to DepEd to address the results that have been found,” said Prof. Rochelle Lucas, chair of the Department of English and Applied Linguistics at DLSU and member of the ‘Research Team.
In addition to providing obvious recommendations such as adding additional resources to schools, especially those in the lower strata, the team also wants to come up with options that can be used to modify certain practices in the school that do not require more time. budget but instead change the student experience. in the way they engage in reading assignments in their classrooms.
The team also plans to create data visualizations to make the results more interesting but informative to the public.
With the amount of data in PISA, Dr Macario Cordel 2nd, Executive Dean of the Data Science Institute, expressed a challenge they faced when analyzing the data.
“If you look at the dataset, you have thousands of variables to consider and then thousands of student data to analyze. It is really difficult for us to target the most important variables or factors that will influence student performance. ,” he said.
The team is currently in the process of training the models better to see if the 81% accuracy can be further improved, as they determine whether or not a certain student profile meets the standards.