High Impact Academic Paper Prediction Using Temporal and Topological Features

Abstract

Predicting promising academic papers is useful for a variety of parties, including researchers, universities, scientific councils, and policymakers. Researchers may benefit from such data to narrow down their reading list and focus on what will be important, and policymakers may use predictions to infer rising fields for a more strategic distribution of resources. This paper proposes a novel technique to predict a paper's future impact (i.e., number of citations) by using temporal and topological features derived from citation networks. We use a behavioral modeling approach in which the temporal change in the number of citations a paper gets is clustered, and new papers are evaluated accordingly. Then, within each cluster, we model the impact prediction as a regression problem where the objective is to predict the number of citations that a paper will get in the near or far future, given the early citation performance of the paper. The results of empirical evaluations on data from several well-known citation databases show that the proposed framework performs significantly better than the state of the art approaches.

Feature Driven and Point Process Approaches for Popularity Prediction

“…Predicting popularity with feature-driven models. Data-driven approaches [11,19] treat popularity as a nondecomposable process and take a bottom-up approach. They rely on machine learning algorithms to make the connection between item popularity and an extensive set of features.…”

“…Similar to the setup in the previous Sec 6.1, we observe cascades for 5 minutes, 10 minutes and 1 hour and fit the Hawkes and Seismic models for each cascade. The train-test split is different, in order to replicate closer the experimental setup in Martin et al [25]: the data from first half of July (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15) is used for training and the data from second half of July (16-31) for testing. We use the News historical data from April to June to construct the past user success feature for the feature-driven approach.…”

Prediction of Company's Trend based on Publication Statistics and Sentiment Analysis

“…There are no existing work related to prediction of company's future trend. Therefore, we compared our method to the most recent state of the art citationbased work related to high impact academic paper prediction proposed by Davletov et al (Davletov et al, 2014). More precisely, we applied Davletov's method (Citation) to open patents and compared it with the result obtained by our method.…”

“…McNamara et al proposed a method for predicting paper's future impact by using topological features extracted from citation network (McNamara et al, 2013). In addition to topological features, Davletov et al predicted high impact academic paper based on temporal features of citations (Davletov et al, 2014). There are a few academic paper prediction method used on textual features (Kogan et al, 2009;Joshi et al, 2010;Yagatama et al, 2011), while much of the previous work on paper prediction used mainly citation statistics (Shi et al, 2010;Yan et al, 2012).…”

A Data-Driven Unified Framework for Predicting Citation Dynamics

“…Hitherto, existing approaches have focused on the prediction of future h-index values at author level [14], [15] or citation counts at publication level [16], [17]. Another categorization of existing approaches occurs with regards to their modeling methodology.…”