Skip to content

pgplarosa/Higher-Education-Research-Dashboard

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 

Repository files navigation

HERD

Higher Education Research Dashboard - Capstone Project

Abstract

Filipinos view research as a positive undertaking that benefits the nation and its needs. However, managing research efforts is difficult because it is systematically disorganized. For the convenience of CHED in developing research policies for the enhancement and sustainability of State Universities and Colleges (SUCs) research thrust, we developed a dashboard that can be used for analytics and reporting with features corresponding to an objective. We conceptualized and implemented a central repository of relevant research data from 11 partner SUCs. Incorporating more data features can aid in increased systemic and efficient utilization of research data. Keywords were imputed using KeyBERT resulting in a 33% increase in imputed keywords. To determine the research priorities of current SUCs and regions, we clustered the research abstracts using HDBSCAN using several clustering metrics and interpreted each cluster as 10 fields of research namely: Education, Crop Cultivation, Data Science, Governance, Pandemic, Utilities, Indigenous People, Livestock Agriculture, Aquaculture, and Biodiversity. The topic distributions of SUCs, regions and funding agencies revealed priority areas in research and can inform research direction such as incentivization, mentorship and multi-disciplinary approaches in addressing UN Sustainable Development Goals (SDGs). We presented four predictive models as utility for the dashboard. School classification is a Linear Support Vector Machine (SVM) model with a precision of 0.46 for recommendation of specialized schools. SDG classification is a Linear SVM with a precision of 0.47 used to measure SUCs’ contribution to SDGs. Topic classification is a Linear SVM with a precision of 0.85 to identify topics for prioritization. Finally, Research similarity using cosine similarity is developed to identify possible collaborations with research authors. To analyze the collaboration dynamics of each SUC, we profiled local co-authorship networks based on several metrics such as node-components ratios, clustering coefficient and average shortest paths resulting in a progressive profile set of (1) highly disconnected networks, (2) presence of small tightly knit or cohesive components and (3) presence of larger developing connected components. University-specific policies can be crafted to improve internal research dynamics based on these collaboration profiles. Centrality and modularity analysis were also performed which identified important authors and research communities for SUC-specific collaboration network analysis. Due to the absence of an inter-university network using co-authorship, a larger network was built involving all authors in the database by using common research topics as connections to find possible inter-university collaboration jump off points. To determine metrics that should be considered in research apart from citation counts, we produced several metrics: (1) tracking research priorities through topic distributions, (2) collaboration through 3 tiers of collaboration profiles, and (3) the impact of research through classification research to their targeted SDGs.

Authors

Mentors

  • Legara, Erika Fille T., PhD
  • Dailisan, Damian N., PhD
  • Dorosan, Michael A., MSc