Living Reference Work Entry

Handbook of Heuristics

pp 1-37

Date: Latest Version

World’s Best Universities and Personalized Rankings

  • Mario Inostroza-PontaAffiliated withDepartamento de Ingeniería Informática, Universidad de Santiago Email author 
  • , Natalie Jane de VriesAffiliated withSchool of Electrical Engineering and Computing, Faculty of Engineering and Built Environment
  • , Pablo MoscatoAffiliated withSchool of Electrical Engineering and Computing, Faculty of Engineering and Built Environment


This chapter presents a heuristic for a multi-objective ranking problem using a dataset of international interest as an example of its application, namely, the ranking of the world’s top educational institutions. The problem of ranking academic institutions is a subject of keen interest for administrators, consumers, and research policy makers. From a mathematical perspective, the proposed heuristic addresses the need for more transparent models and associated methods related to the problem of identifying sound relative rankings of objects with multiple attributes. The low complexity of the method allows software implementations that scale well for thousands of objects as well as permitting reasonable visualization. It is shown that a simple and multi-objective-aware ranking system can easily be implemented, which naturally leads to intuitive research policies resulting from varying scenarios presented within. The only assumption that this method relies on is the ability to sort the candidate objects according to each given attribute. Thus the attributes could be numerical or ordinal in nature. This helps to avoid the selection of an ad hoc single score based on an arbitrary assignment of attributes’ weights as other heuristics do. To illustrate the use of this proposed methodology, results are presented and obtained using the dataset on the ranking of world universities (of the years 2007–2012), by academic performance, published annually by ARWU.


Analytics Digital humanities Pareto optimality Ranking Symbolic regression