Citation Count Prediction Based on Academic Network Features

Abstract

Citation count is an important factor to measure the influence of academic publications. Identifying future citation count in advance can help scientists to find references and research area. There are many academic network features which are related to citation count. However, these features have not been completely explored in the existing studies. In this paper, we propose a citation count prediction model based on academic network features. Firstly, some important features are introduced and analyzed in detail. Then, we verify the importance of each feature and use a neural network model to select a set of optimal features. Finally, we present several machine learning methods and one multiple linear regression strategy to predict a paper's future citation. Experimental results on real datasets demonstrate that our model significantly outperforms the baseline method.