Prediction of future citation count with machine learning and neural network

Prediction of long-term citation counts of paper is increasingly important in academic community. In this work, the prediction of future citation counts is considered as binary classification task. The study is based on 2600 papers of physiology extracted from Web of Science. Eight bibliometric features of citing paper in the first three years after the publication of paper were selected. Three machine learning model and a neural network are built to test whether these features are effective in prediction of future citation counts. The experimental result shows that the selected features are valuable in the prediction of long-term citation counts. And the machine learning and neural network are useful for the task of predicating future citation counts.