Feature Characterization in Some Weighted Directed Networks

Abstract

The paper studies the time evolution of weighted directed networks by evaluating some of their fundamental features along time. The main concepts are illustrated by analyzing the evolution of a network which represents the call flow between the towers of a telephone operator in Senegal. Different time scales (hour, day, week, month) are considered to compute the evolution of features in order to characterize properties which may follow cyclic behaviors. A joint characterization of the features is also performed via a paired correlation analysis which also shows to be sensitive to time scales, suggesting different seasonal behaviors in the communication patterns.