Contextual personalized recommendation technique for academic event selection / Parukutty Raman

• 主要作者: Parukutty , Raman
• 格式: Thesis
• 出版: 2018
• 主题: QA75 Electronic computers. Computer science

A recommender system in the academic domain has significant importance to researchers. Incorporating contextual information in a recommender system is vital to ensure that the recommended information is relevant and in accordance to the user’s preferences. Consideration of “context” in the selection of an academic event is likely to have a profound effect on retrieving better recommendation results. Academic events can be classified as events related to any academic domain discipline, such as workshop, seminar or conference. A Recommendation System (RS), using a classical filtering approach, tends to fail when insufficient user preference information is available. The two main classical approaches are content based and collaborative. The objective of this study is to identify the most important contexts in selecting an academic event and develop a contextual personalized recommender technique for academic event selection. A survey is conducted to identify the most important contexts. Four important contexts of time, schedule, location and cost were identified and used in the technique development. Next, a tool was developed using context pre-filtering and collaborative searching techniques. The context will be pre-filtered and a search input that carries contextual data and keywords will be used to search for relevant events using a match matrix from the event database. The same events will then be checked again to see whether they have been attended by any neighbour of the user using the Top N weighted nearest neighbour technique. Contexts and keywords are explicitly given by the user. Average precision and mean average precision are used to evaluate the tool. The results will show that contextual personalized event selection techniques produce more relevant results than a technique that only uses classical approaches. This study proposes a context based personalized recommender technique to assist researchers in finding relevant academic events. The developed technique can also be used in other domains.