A Global Flight Networks Analysis Approach Using Markov Clustering and PageRank

The analysis of the global flight network is of significance in many academic fields including social and political science. Existing studies restrict their scope to individual countries, and many of them conduct their researches over flight lengths between airports, which may lose important information from graph-based datasets, such as flight networks. This paper applies Markov Clustering Algorithm based on the idea of random walk to analyse the global flight network and PageRank to research the details of the airports in each cluster. The results demonstrated the power of machine learning algorithms to reveal naturally-formed regions bounded by airline flights that go beyond borderlines of countries, and this research also make an interesting discovery that airports which play important roles of connecting multiple regions (e.g. JFK) are not necessarily the barycenter of a local flight network.