Event Prediction in Complex Social Graphs using One-Dimensional Convolutional
Neural Network
Abstract
Social network graphs possess apparent and latent
knowledge about their respective actors and links
which may be exploited, using effective and effi-
cient techniques, for predicting events within the
social graphs. Understanding the intrinsic relationship patterns among spatial social actors and their
respective properties are crucial factors to be taken
into consideration in event prediction within social
networks. My research work proposes a unique approach for predicting events in social networks by
learning the context of each actor/vertex using neighboring actors in a given social graph with the goal
of generating vector-space embeddings for each vertex. Our methodology introduces a pre-convolution
layer which is essentially a set of feature-extraction
operations aimed at reducing the graph’s dimensionality to aid knowledge extraction from its complex
structure. Consequently, the low-dimensional node
embeddings are introduced as input features to a
one-dimensional ConvNet model for event prediction about the given social graph. Training and evaluation of this proposed approach have been done
on datasets (compiled: November, 2017) extracted
from real world social networks with respect to 3
European countries. Each dataset comprises an average of 280,000 links and 48,000 actors