Studies on MapReduce based Area Skyline Query and Parameter Estimation of Queueing Systems
Chen Li · Hiroshima University Acedemic Information Repository (Hiroshima University) · 2019
In this dissertation, the author discusses two kinds of research works, MapReduce based computation of area skyline query for location recommendation, and parameter estimation of queueing systems with utilization data, respectively.With the rapid development of computer technology and network technology, a vast amount of available information may cause information overload problem.The exponential growth of data makes customers inundate with various choices.Since the ability to resolve the ``information overload'' problem, recommendation systems are widely accepted in E-commerce in recent years.The basic premise of recommenders is to reduce noise and filter out information, which is not relevant to customers' tastes.In the past decades, many researchers worked on generating good recommenders.Recommendation systems can help enterprises improve economies.For example, Amazon.comclaimed that product sales grew 35% from the recommendation system.Moreover, Netflix.comclaimed that 66% of movies were rented by using recommendation systems.Google News Recommendations generated 38% more click-throughs.Location recommendations are the main part of recommendation systems.For example, the selection of good locations is essential in the service filed, which may help customers find implicit preferable places in an unfamiliar environment.Moreover, it is also critical for a businessperson to find a suitable location to build a company.Skyline query is a well-known information retrieval approach for recommendation systems.Skyline query can select a small number of representative objects from an extensive database, which already has been applied in the location selection problem.For example, a traveler wants to find a hotel from several candidate hotels.In this case, we can use the skyline query to help the traveler select a preferable hotel.However, in some real-world scenarios, the candidate points are not given for the