Using Deep Learning for Temporal Forecasting of User Activity on Social Media: Challenges and Limitations

The recent advances in neural network-based machine learning algorithms promise a revolution in prediction-based tasks in a variety of domains. Of these, forecasting user activity in social media is particularly relevant for problems such as modeling and predicting information diffusion and designing intervention techniques to mitigate disinformation campaigns. Social media seems an ideal context for applying neural network techniques, as they provide large datasets and challenging prediction objectives. Yet, our experiments find a number of limitations in the power of deep neural networks and traditional machine learning approaches in predicting user activity on social media platforms. These limitations are related to dataset characteristics due to temporal aspects of user behavior. This work describes the challenges we encountered while attempting to forecast user activity on two popular social interaction sites: Twitter and GitHub.

[1]  Zhenyu Wang,et al.  Cascade2vec: Learning Dynamic Cascade Representation by Recurrent Graph Neural Networks , 2019, IEEE Access.

[2]  Max Welling,et al.  Semi-Supervised Classification with Graph Convolutional Networks , 2016, ICLR.

[3]  Cyrus Shahabi,et al.  Diffusion Convolutional Recurrent Neural Network: Data-Driven Traffic Forecasting , 2017, ICLR.

[4]  Jean-Loup Guillaume,et al.  Fast unfolding of communities in large networks , 2008, 0803.0476.

[5]  K. Koch Introduction to Bayesian Statistics , 2007 .

[6]  Albert-László Barabási,et al.  The origin of bursts and heavy tails in human dynamics , 2005, Nature.

[7]  Philip S. Yu,et al.  A Comprehensive Survey on Graph Neural Networks , 2019, IEEE Transactions on Neural Networks and Learning Systems.

[8]  Quoc V. Le,et al.  Sequence to Sequence Learning with Neural Networks , 2014, NIPS.

[9]  Hui Li,et al.  A Deep Learning Approach to Link Prediction in Dynamic Networks , 2014, SDM.

[10]  Xavier Bresson,et al.  Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering , 2016, NIPS.

[11]  Ali Farhadi,et al.  You Only Look Once: Unified, Real-Time Object Detection , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[12]  Kate Starbird,et al.  Disinformation as Collaborative Work , 2019, Proc. ACM Hum. Comput. Interact..

[13]  Kaiming He,et al.  Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks , 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[14]  Jure Leskovec,et al.  node2vec: Scalable Feature Learning for Networks , 2016, KDD.

[15]  Cao Xiao,et al.  FastGCN: Fast Learning with Graph Convolutional Networks via Importance Sampling , 2018, ICLR.

[16]  Krishna P. Gummadi,et al.  Measuring User Influence in Twitter: The Million Follower Fallacy , 2010, ICWSM.

[17]  Zhanxing Zhu,et al.  Spatio-temporal Graph Convolutional Neural Network: A Deep Learning Framework for Traffic Forecasting , 2017, IJCAI.

[18]  Yoshua Bengio,et al.  Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling , 2014, ArXiv.

[19]  Samy Bengio,et al.  Scheduled Sampling for Sequence Prediction with Recurrent Neural Networks , 2015, NIPS.

[20]  Zhiyuan Liu,et al.  Graph Neural Networks: A Review of Methods and Applications , 2018, AI Open.

[21]  Jure Leskovec,et al.  SEISMIC: A Self-Exciting Point Process Model for Predicting Tweet Popularity , 2015, KDD.

[22]  Jimmy Ba,et al.  Adam: A Method for Stochastic Optimization , 2014, ICLR.