A web based document archiving system using indexing and machine learning for research and innovation grant allocation.
Date
2024
Authors
Lupyani, Rebecca
Journal Title
Journal ISSN
Volume Title
Publisher
The University of Zambia
Abstract
In today’s rapidly evolving world, research forms a cornerstone of human progress leading to new products, services, and technologies, which, in turn, can stimulate economic growth and enhance the quality of life. Research enables people to learn, innovate and address the complex challenges facing a society. It is a powerful tool for making the world a better place and as such many countries endeavor to support the research landscape by providing research grants in different sectors. The process of allocating research grants plays a pivotal role in fostering scientific progress, innovation and knowledge. The traditional manual selection of grant proposals, while well-established can be resource intensive time-consuming, subjective, and
prone to bias. This paper presents an unconventional strategy that leverages machine learning algorithms to enhance the fairness, efficiency, and transparency of the grant allocation process by removing human biases and prejudices that can inadvertently influence funding decisions. The study discussed the design and implementation of a machine learning-based grant allocation system using historical grant data from a reputable funding agency and provided empirical evidence of its effectiveness by selecting the best performing text classification algorithm from a comparative analysis of three models and integrating it into a web based application. The three models compared were the K- Nearest Neighbour, the Naives Bayes and the Support Vector Machine. The Support Vector Machine exhibited the highest performance metrics with an accuracy percentage of 88%, precision of 86%, recall of 87% and F1 score of 87%, whereas the K-Nearest Neighbour portrayed the lowest performance with an accuracy of 41%, precision 52%, recall of 43% and F1 score of 37%. Thus, the Support Vector Machine was integrated into a web based application to facilitate fund allocation. The developed system will promote fair review of research and innovation proposal applications by automatically categorizing them into topic categories that facilitate funding such as engineering, science and technology. The system will also assist in tracking and monitoring the progress for the research projects for which funding institutions invest in and also in digitally archiving the documents in an effective and efficient manner
Description
Thesis of Master of Science in Computer Science.