Link Mining and Temporal Link Prediction

Abstract

The ability to predict links among data objects is central to many data mining tasks such as social network, business analytics, and recommendation system. A substantial literature has adapted static graph presentation where a snapshot of network is analyzed to predict hidden or future links, those are based on node wise similaritydistance based approaches or topological pattern based methods. Further, forecasting methods to setup reasonable expectations for future links (events) are mainly based on properties of relevant historical data (timeseries) these model expresses relationship between what the past evidence about link in objects shows and what could possibly take place in future. In general, link prediction problem is modeled as, given link data for times 1 through T, the task is to predict links at time T+1, And if data has underlying periodic structure, upto what extend predictions can be made in future time T+2,T+3………T+k.(k>0) Current approaches do not explicitly handle dynamic user preferences and thus proposed work is contributed for development of interactive framework that considers temporal evolutions of link occurrences to predict and forecast link occurrence probabilities at particular time in future. General Terms: Link mining, Link prediction, Forecasting, predictive function