They find factors affecting students that are often overlooked in education reform
Earlier this month, the Department of Education (DepEd) called out the World Bank (WB) for basing their report on 2018 and 2019 data from three international assessments and for not even acknowledging the department’s reform initiatives for quality education even before their participation.
It’s no secret that Philippine basic education has been reforming itself time and again. In fact, the previous and present governments have already applied different policies, plans, and programs that were supported by huge funds and million dollar loans.
Many would agree that the move by the present administration to participate in international assessments such as Programme for International Student Assessment (PISA) could really gauge the effectiveness of the reforms and could also establish our baseline in relation to global standards.
But how can we really benefit from the large results’ data of these international assessments? For a team of data scientists from Dr. Andrew L. Tan Data Science Institute (ALTDSI) of De La Salle University (DLSU), a deeper data mining and analysis is needed to determine not just problems but also their root causes that have been affecting the performance of students. And they are doing it through machine learning (ML), which utilizes computer systems, algorithms, and statistical models in analyzing and drawing inferences from patterns in data.
Using the computer models they trained, the scientists are also able to predict which students can meet or may fall below proficiency standards and what are the factors that categorize individuals who are vulnerable to be in the poor performing group.
The 2018 PISA data on reading, where 15-year old Filipino students ranked last, was their initial subject for the ML approach. If DepEd considers PISA 2018 results outdated, DLSU data scientists, on the other hand, see them as a mine where gems of information could be dug up and used to improve the students’ performance level.
The research project engaged members from different disciplines to sensibly examine the PISA data set. Among them are NAST Academician and Psychology Department professor Allan Bernardo, ALTDSI Dean Dr. Macario Cordel II, Chair of the Department of English and Applied Linguistics professor Rochelle Irene Lucas, and Software Technology faculty members and ALTDSI faculty affiliates Unisse Chua, Sashmir Yap, and Jude Teves.
With the enormous amount of data from PISA, Dr. Macario said that it was really challenging for them to zero in the most important variables or factors that will influence the student performance.
“But with machine learning models, you could model a more complex nonlinear system. And using its learning algorithm, it could update its values as we feed more and more data. So we are confident that we can get a better classification result and do more reliable analysis in terms of which features are important,” beams Dr. Mac.
As of the moment, they are training the models better to increase their accuracy. Jude declared that they are still trying to add more sophisticated models. Currently, with the models that they have used, they are converging at 81-82 accuracy. They’re hoping that it will still improve by using more advanced techniques.
According to professor Allan, their preliminary findings that present factors why students fall below proficiency standard overlap with what the WB instated in its report such as bullying, negative belief or mindset, and low socioeconomic status.
“However, our study using ML provides a more complex understanding that also points out what contributes to those problems. So we work with factors within the students’ environment that influence their beliefs, motivations, aspirations, and experiences, which are then often overlooked in the reform efforts.” explains professor Allan.
“Addressing these factors doesn’t really need massive investment.” he continues. “We can come up with options that school heads, teachers, and community leaders can do within the students’ environment that are not necessarily dependent on how much budget the government allocates.”
The team will present their study and findings with policy makers, including DepEd. “We are preparing the policy brief, and that will definitely include our recommendations to DepEd to address the results that were found,” says professor Rochelle.
They are also developing data visualization of the findings that are most useful so that it would be easier for policy makers, school heads, and education stakeholders to understand and find interesting yet informative.
The team is confident that exploring such factors using ML, their findings and recommendations won’t leave unnoticed for the country’s education reform.