Automating Conference Paper Assignment Using Classification Algorithms Incorporated with TF-IDF Vectorisation

Assign submitted papers to the right topics or reviewers with the right expertise in a research conference is important. However, the paper assignment process is presently a repetitive and time-consuming manual task. This study aims to automate the process by building a classification model incorporating TF-IDF vectorisation. We outline a data mining methodology consisting of data selection, text pre-processing, transformation, classification, evaluation, and optimisation. A data set of 1,000 conference papers with five research topics was carefully selected, pre-processed, and transformed into a TF-IDF vector representation. The TF-IDF vector was then applied to six popular document classification algorithms: Support Vector Machine (SVM), Naïve Bayes (NB), Logistic Regression (LR), K-Nearest Neighbour (KNN), Decision Tree (DT) and Random Forest (RF). Initially, these algorithms with default parameters were evaluated and measured using F-measure. Then, the top three among them were further optimised using Grid Search. The results revealed SVM as the best-performing algorithm, with an F-measure score of 0.907.