Research literacy and its predictors among university students and graduates identified by machine learning and spatial analysis

TL;DR

This study explores research literacy among university students and graduates, finding that many lack it, and identifies factors like research training and professional engagement as important predictors.

Contribution

This is the first study to comprehensively assess research literacy using GIS and machine learning alongside traditional statistical methods.

Findings

Over half of the participants had poor research literacy.

Taking research courses outside university was the most impactful factor for higher research literacy.

Random Forest and CatBoost models achieved high accuracy in predicting research literacy.

Abstract

The landscape of academic publishing has evolved dramatically, leading to a surge in publications and journals. The ‘publish or perish’ culture has resulted in undesirable practices, such as many researchers publishing in predatory journals due to institutional pressures and lack of awareness. While numerous studies have investigated knowledge of predatory journals, overall research literacy has remained underexplored. This study is the first to assess research literacy comprehensively, incorporating GIS and machine learning techniques alongside traditional statistical analyses. This study utilized a cross-sectional survey method with a questionnaire collecting information on socio-demographics, academic information, research training and experience, and research literacy. Traditional statistical analyses were performed using SPSS, while machine learning models were developed with Python and Google Colab. Supervised classification algorithms and mapping with R statistical software’s ‘bangladesh’ package. The findings revealed that over half of the participants had poor research literacy. Significant predictors of higher research literacy included satisfaction with research courses at university education, research course taken outside university , and research-related professional engagement. Machine learning analysis identified that taking research courses outside of university was the most impactful factor for research literacy, while researchers within family members had minimal influence. The Random Forest and CatBoost models performed strongly in predicting literacy, achieving accuracy rates of 73.04% and 71.57%, respectively, and precision values of 73.29% and 71.69%, respectively, with low log loss values of 0.57 and 0.56. GIS-based spatial analyses revealed regional disparities in research literacy (χ²=9.234, p = 0.236), with certain divisions exhibiting a higher prevalence of lower literacy. This study highlights that a substantial portion of the participants lack research literacy, which is associated with multiple factors. The findings suggest the need for intervention programs to enhance research practices and awareness among students and professionals, fostering a culture of academic excellence.

The online version contains supplementary material available at 10.1038/s41598-025-19488-4.

Figures