Understanding Acdamic Impact Development by Predicting the G-index In Collaboration Networks

Academic impact is an important factor in the research assessment of a scholar. The prediction of the individual future academic impact is beneficial to detecting talented scholars and allocating the academic resources more reasonably. Recently, many evaluation indicators based on academic impact have been proposed, among which the G-index is more objective than the citations and H-index that used more frequently. At present, the main contradiction is formed between the great values of the individual future academic impact in practice and the imperfect evaluation and prediction methods. In order to solve this contradiction, this paper chooses the objective and comprehensive evaluation indicator G-index to measure individual academic impact and proposes the solution of how to predict the G-index at a given time interval for the first time in relevant research fields. Firstly, different characteristic features are extracted based on the structure of the academic social network. Then, a prediction model based on deep learning is proposed to learn the function mapping relationship between the future G-index time with the current feature set. Finally, the proposed model is validated with real academic social network dataset. The experiment results show that the individual G-index development is tightly related to the scholar's characteristic features in the current academic social network and the deep learning model performs well in predicting the future G index.