Year of Graduation
Link Direction Prediction in Semantic Networks
Mathematical Methods of Modelling and Computer Technologies
This work solves the link direction prediction in semantic networks problem. Associations are considered in the terms of associations between lexical concepts derived from Wordnet database. The influence of distributive semantic based features and graph based features on the results of binary classification (direct link, lack of link) and four class classification (outbound link, inbound link, bidirectional link, lack of link) was studied. Classification accuracy of the models achieved 89.3% and 84.1% for binary and four class classifications respectively.