Data Mining Using Neural Network for Research Topic Classification Based on Institutional Research Roadmap

Abstract

In the last five years, lecturers and students at the Faculty of Medicine at the Islamic University of Bandung have produced hundreds of studies. Still, studies on the suitability of various studies with topics according to the institution's research roadmap have not been carried out. This research aimed to classify research documents based on 12 research roadmap topics. The data used in this study are the research titles of lecturers and students in 2015-2021, amounting to 1064 data. The document extraction process uses text mining, while the document grouping process is carried out using a supervised method with an artificial neural network algorithm. At the text mining stage, preprocessing procedures are carried out in case folding, tokenization, and filtering, followed by analysis through weighting using IDF and evaluating accuracy, precision, and recall. The Neural Network Algorithm can classify by level. The classification results using the neural network algorithm show that overall from the training data, the average precision is 74.7%, recall is 74.3%, and accuracy is 74.3%. Of the 12 research topics, the training dataset obtained high accuracy, precision, and recall values found on herbal medicine, Islamic insert, industrial health, tuberculosis, and vaccination. The four topics are in line with the five institution's leading research topics. The results of dataset testing analysis found that the research topics carried out by students and lecturers of Unisba School of Medicine were distributed among the 12 research topics on the institutional roadmap with the increased trends in institutional leading research topics.