Cold Start Problem Resolution Using Bayes Theorem

摘要

With the plethora of available data online, academicians and researchers find it difficult to retrieve the relevant data. The problem complexity increases for the new academician and the new paper added with no previous knowledge of the type of useful research papers and no acknowledgment of their new work done. This is inferred as the cold start problem in recommender system having few or near zero ratings. This problem enables us to propose the methodology wherein we could able to provide the ratings of each work done of the academicians using collaborative filtering to develop hidden relation between the titled paper and its corresponding references and citations. The rating is provided using Bayes theorem conditional probability for each research paper. Hence, a new academician in research area will be recommended based on the rating in the form of calculated probability. Besides, it also sufficiently resolves the issue of sparsity, which presents the low or no ratings of the research work done. Other issues such as researcher details, probable attacks of fake papers and trust bounded feedback are also catered in the requisite study. Further, the study is labeled based on set threshold to recommend or not.