Tailored feedforward artificial neural network based link prediction

Abstract

Link prediction is about reckoning the probability of a link existence between two nodes in any network. A network can also be observed as a graph structure. Paper uses network as graph structure here. Deep learning is the driving force behind the current machine learning success and also achieved the state of art performance in various domains. Purpose of the paper here is to obtain link prediction using deep feed forward artificial neural network (DFFANN) which is one of the deep learning technique. Here deep artificial neural network and the proposed Tailored deep feed forward artificial neural network (TDFFANN) suitability is explored for link prediction. Proposed TDFFANN on same datasets gives improved performance as compared to other ANN. Results clearly shows that, the deep learning when used with optimized input and efficient evaluation methods make better predictions. Hence TDFFANN is the need of time to get more accurate link predictions. In general TDFFANN for link prediction out perform simple ANN and Simple deep ANN on the datasets used.