Track: Artificial Intelligence

Abstract

This work is aimed at building prediction model that leverages data sources from the ranking body to estimate ranking position, given relevant parameters. Historical data of previous rankings from 2012 to 2016 were obtained from the database of Times Higher Education (THE). A single-layered, feed-forward Artificial Neural Network (ANN) with varying neurons (1-10) was set up based on the Scaled Conjugate Gradient (SCG) back-propagation algorithm. Ranking data that were based on Thomson Reuter’s Web of Science database (2012-2015) were used for training the ANN. Model validation and testing were conducted with Scopus database ranking data (2016). The fitness and accuracy of the prediction model was evaluated based on statistical analysis. Statistical analysis showed that the prediction model gives a good and acceptable performance despite the recent change in the choice of database by the ranking body. The overall regression co-efficient for the network training was 0.964. A validation performance of 95.23% showed that the prediction model generalizes well enough for data that were not included in the training data set. The Mean Absolute Error (MAE), the Root Mean Square Error (RMSE), and the Standard Deviation (SD) were 5.59, 7.65, and 5.23 respectively. Therefore, the model guarantees acceptable prediction results.