PROJECT TITLE :

Predicting Hot Events in the Early Period through Bayesian Model for Social Networks

ABSTRACT:

It is essential for a wide variety of applications, such as information dissemination mining, ad recommendation, and others, to make accurate predictions of emerging hot events at an early stage. The currently available methods either require an extended period of observation over the occurrence or the extraction of features that are difficult and costly to accomplish. However, because there is a lack of data at this early stage of an emerging event, the temporal characteristics of hot events and non-hot events are not yet sufficiently distinguishable from one another. We present BEEP, a Bayesian perspective Early stage Event Prediction model, as a solution to this conundrum within the context of this body of work. We formulate the problem of predicting hot events using two Semi-Naive Bayes Classifiers. In this formulation, we take into account both the structural and temporal features of the data and then perform a distribution test on the features that have been selected. Our approaches have been shown to be successful through both theoretical investigation and extensive empirical testing on two different real-world datasets.