Research Article

An Iterative Information-Theoretic Approach to the Detection of Structures in Complex Systems

Figure 3

(a) The temporal behaviour of the 136 variables over the 101 points-of-control, which are distributed over time without a regular frequency. The involved stakeholders’ attendance at the points-of-control of the GC project is marked in black (the absence is marked in white). (b) The same variables resorted in order to highlight the RSs found by the RI algorithm. The two coloured bars highlight (i) the different “final” groups (lower coloured bar) and (ii) the internal subdivisions of the final groups (the smallest RSs detected during the 26 iterations of the sieving procedure). Group G was obtained by grouping smaller RSs: for easier graphical representation, the exact multistep sequence of the assembly has not been represented. The same observations hold for the assembly processes in general: for example, the steps leading to grouping RSs B, C, E, G1, G2, and G3 are intertwined, while groups G4 and G3 are formed “before” the final expression of group G1. The other variables (termed as the “remaining part”) could not be assigned to any group with a sufficiently high degree of significance.