AN OPTIMAL HYPER-PARAMETER TUNED TRANSFER LEARNING AIDED SCHEME FOR SCIENTIFIC PAPERCITATION RECOMMENDATION SYSTEM

Background: In recent times, Recommendation system (RS) shows a vibrant part in both industries and research.In academia field, there exists a need and urge to aid researchers for discovering suitable and appropriate scientific information's over recommendations.However, there is huge gap among existing RS and the real-time issues.For overcoming this existing issue, and aids research scholars to choose effective citation, a new transfer learningbased model is proposed for the citation recommendation system.Methodology: At first, input dataset that are publicly accessible are considered and removal of stop words or functions words will be carried with the use of stochastic pattern matching approach through which the creation of library is carried.After that, the document canonicalization is performed with the use of Term Frequency-Inverse Document Frequency (TF-IDF) after which the stored data are predicted and classified by means of Improved Pre-trained Inception V3 classifier. Before classification process, the hyper-parameters are tuned on employing Multi-parameter Hunter prey optimization approach (MP-HPO). This in turn enhances the classifier performance aiding in effective recommendation of citation records. At last, ranking is done by means of employing weighted Pearson correlation coefficient to rank the predicted citation and recommend best ones by means of ranking strategy.Result and discussion: The proposed technique designed is validated on various metrics like precision, Mean Average Precision (MAP), Mean Reciprocal Rank (MRR), Recall, and F1-score.The proposed outcomes are compared with various existing schemes to show the enhancement of proposed strategy over others.Conclusion: The validated outcome reveals the superiority of proposed model over other existing models compared.