Limits...
Failure dynamics of the global risk network.

Szymanski BK, Lin X, Asztalos A, Sreenivasan S - Sci Rep (2015)

Bottom Line: Using the maximum likelihood estimation, we find the optimal model parameters and demonstrate that the model including network effects significantly outperforms the others, uncovering full value of the expert collected data.Our findings include such risk properties as contagion potential, persistence, roles in cascades of failures and the identity of risks most detrimental to system stability.The model provides quantitative means for measuring the adverse effects of risk interdependencies and the materialization of risks in the network.

View Article: PubMed Central - PubMed

Affiliation: 1] Social and Cognitive Networks Academic Research Center, Rensselaer Polytechnic Institute, Troy NY 12180 [2] Dept. of Computer Science, RPI, 110 8th Street, Troy, NY 12180 [3] Dept. of Computer Science &Management, Wroclaw University of Technology, 50-370 Wroclaw, Poland.

ABSTRACT
Risks threatening modern societies form an intricately interconnected network that often underlies crisis situations. Yet, little is known about how risk materializations in distinct domains influence each other. Here we present an approach in which expert assessments of likelihoods and influence of risks underlie a quantitative model of the global risk network dynamics. The modeled risks range from environmental to economic and technological, and include difficult to quantify risks, such as geo-political and social. Using the maximum likelihood estimation, we find the optimal model parameters and demonstrate that the model including network effects significantly outperforms the others, uncovering full value of the expert collected data. We analyze the model dynamics and study its resilience and stability. Our findings include such risk properties as contagion potential, persistence, roles in cascades of failures and the identity of risks most detrimental to system stability. The model provides quantitative means for measuring the adverse effects of risk interdependencies and the materialization of risks in the network.

No MeSH data available.


Related in: MedlinePlus

Persistence and materialization probabilities of risks in cascades:(a) The bar graph shows the fraction of the total lifetime of a cascade that a given risk is expected to be active, as obtained from 106 simulations for each of 15 different initiators, where initiators are chosen from sets of risks with high contagion potential, medium contagion potential and low contagion potential. The specific risks chosen as initiators were risks 1,8,9,12,16,20,23,25,26,27,31,33,42,47,49. (b) The bar graphs show the materialization probabilities of four labeled risks, as a function of the initiator of the cascade (whose names are shown within their respective bars). Each experiment ended when either the selected risk was infected, or all risks became inactive.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4471900&req=5

f5: Persistence and materialization probabilities of risks in cascades:(a) The bar graph shows the fraction of the total lifetime of a cascade that a given risk is expected to be active, as obtained from 106 simulations for each of 15 different initiators, where initiators are chosen from sets of risks with high contagion potential, medium contagion potential and low contagion potential. The specific risks chosen as initiators were risks 1,8,9,12,16,20,23,25,26,27,31,33,42,47,49. (b) The bar graphs show the materialization probabilities of four labeled risks, as a function of the initiator of the cascade (whose names are shown within their respective bars). Each experiment ended when either the selected risk was infected, or all risks became inactive.

Mentions: Next, we investigate which risks are predominantly responsible for the cascades persisting for such long time-scales. Fig. 5 shows the expected fraction of the lifetime of a cascade for which a particular risk is active, in ranked order. The top five highest active risks are 8, active for 83% of the cascade lifetime, 1, active for 53% of the lifetime, 17, active for 46% of the lifetime, 40, active for 39% of the lifetime, and 12, active for 35% of the lifetime. Interestingly, the lists of top five most persistent risks observed in the cascades and seen in the full dynamics of activation (when all nodes undergo both internal and external activation Poisson processes) are identical.


Failure dynamics of the global risk network.

Szymanski BK, Lin X, Asztalos A, Sreenivasan S - Sci Rep (2015)

Persistence and materialization probabilities of risks in cascades:(a) The bar graph shows the fraction of the total lifetime of a cascade that a given risk is expected to be active, as obtained from 106 simulations for each of 15 different initiators, where initiators are chosen from sets of risks with high contagion potential, medium contagion potential and low contagion potential. The specific risks chosen as initiators were risks 1,8,9,12,16,20,23,25,26,27,31,33,42,47,49. (b) The bar graphs show the materialization probabilities of four labeled risks, as a function of the initiator of the cascade (whose names are shown within their respective bars). Each experiment ended when either the selected risk was infected, or all risks became inactive.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4471900&req=5

f5: Persistence and materialization probabilities of risks in cascades:(a) The bar graph shows the fraction of the total lifetime of a cascade that a given risk is expected to be active, as obtained from 106 simulations for each of 15 different initiators, where initiators are chosen from sets of risks with high contagion potential, medium contagion potential and low contagion potential. The specific risks chosen as initiators were risks 1,8,9,12,16,20,23,25,26,27,31,33,42,47,49. (b) The bar graphs show the materialization probabilities of four labeled risks, as a function of the initiator of the cascade (whose names are shown within their respective bars). Each experiment ended when either the selected risk was infected, or all risks became inactive.
Mentions: Next, we investigate which risks are predominantly responsible for the cascades persisting for such long time-scales. Fig. 5 shows the expected fraction of the lifetime of a cascade for which a particular risk is active, in ranked order. The top five highest active risks are 8, active for 83% of the cascade lifetime, 1, active for 53% of the lifetime, 17, active for 46% of the lifetime, 40, active for 39% of the lifetime, and 12, active for 35% of the lifetime. Interestingly, the lists of top five most persistent risks observed in the cascades and seen in the full dynamics of activation (when all nodes undergo both internal and external activation Poisson processes) are identical.

Bottom Line: Using the maximum likelihood estimation, we find the optimal model parameters and demonstrate that the model including network effects significantly outperforms the others, uncovering full value of the expert collected data.Our findings include such risk properties as contagion potential, persistence, roles in cascades of failures and the identity of risks most detrimental to system stability.The model provides quantitative means for measuring the adverse effects of risk interdependencies and the materialization of risks in the network.

View Article: PubMed Central - PubMed

Affiliation: 1] Social and Cognitive Networks Academic Research Center, Rensselaer Polytechnic Institute, Troy NY 12180 [2] Dept. of Computer Science, RPI, 110 8th Street, Troy, NY 12180 [3] Dept. of Computer Science &Management, Wroclaw University of Technology, 50-370 Wroclaw, Poland.

ABSTRACT
Risks threatening modern societies form an intricately interconnected network that often underlies crisis situations. Yet, little is known about how risk materializations in distinct domains influence each other. Here we present an approach in which expert assessments of likelihoods and influence of risks underlie a quantitative model of the global risk network dynamics. The modeled risks range from environmental to economic and technological, and include difficult to quantify risks, such as geo-political and social. Using the maximum likelihood estimation, we find the optimal model parameters and demonstrate that the model including network effects significantly outperforms the others, uncovering full value of the expert collected data. We analyze the model dynamics and study its resilience and stability. Our findings include such risk properties as contagion potential, persistence, roles in cascades of failures and the identity of risks most detrimental to system stability. The model provides quantitative means for measuring the adverse effects of risk interdependencies and the materialization of risks in the network.

No MeSH data available.


Related in: MedlinePlus