Application of Data Mining Techniques for Article Evaluation Results Prediction from Journal of Social Innovation and Lifelong Learning

Abstract

This research aims to (1) develop a predictive model for article evaluation results based on data from the Journal of Social Innovation and Lifelong Learning, and (2) compare the performance of data mining techniques used. The study analyzed evaluation data from 170 articles using Naive Bayes, Decision Tree, and K-Nearest Neighbors techniques implemented in Python on the Google Colab platform. Feature selection using Information Gain from 19 attributes identified 10 key features important for prediction: language use, table and illustration formats, article and supplementary file quality, clarity and detail of the abstract and summary, specification of research objectives in the abstract and summary, identification of research methods in the abstract and summary, research results in the abstract and summary, clarity of research questions, research methods, and discussion of research findings. The predictive models were then built using Naive Bayes, Decision Tree, and K-Nearest Neighbors. The results showed that Naive Bayes had the lowest accuracy at 20.66%, while K-Nearest Neighbors achieved the highest overall accuracy at 82.46%. However, the Decision Tree produced more consistent results. The findings of this research can be applied to develop a support system for article evaluation and serve as a guideline for authors to improve article quality, thereby increasing their chances of publication