A heuristic optimization method for mitigating the impact of a virus attack

Taking precautions before or during the start of a virus outbreak can heavily reduce the number of infected. The question which individuals should be immunized in order to mitigate the impact of the virus on the rest of population has received quite some attention in the literature. The dynamics of the of a virus spread through a population is often represented as information spread over a complex network. The strategies commonly proposed to determine which nodes are to be selected for immunization often involve only one centrality measure at a time, while often the topology of the network seems to suggest that a single metric is insufficient to capture the influence of a node entirely. In this work we present a generic method based on a genetic algorithm (GA) which does not rely explicitly on any centrality measures during its search but only exploits this type of information to narrow the search space. The fitness of an individual is defined as the estimated expected number of infections of a virus following SIR dynamics. The proposed method is evaluated on two contact networks: the Goodreau’s Faux Mesa high school and the US air transportation network. The GA method manages to outperform the most common strategies based on a single metric for the air transportation network and its performance is comparable with the best performing strategy for the high school network.

💡 Research Summary

The paper addresses the problem of selecting a limited set of nodes to immunize in order to minimize the spread of a virus on complex networks. While most prior work relies on a single centrality measure—such as degree, betweenness, or eigenvector—to rank nodes, the authors argue that a single metric cannot fully capture a node’s influence, especially in heterogeneous topologies. To overcome this limitation, they propose a genetic algorithm (GA) that searches for near‑optimal immunization sets while using multiple centrality scores only to restrict the search space.

First, the three centralities are computed for every vertex. Nodes are sorted in descending order for each metric, and the top l nodes from each ranking are retained. The union of these three truncated lists, denoted R₀, forms the candidate pool. This reduction is based on the assumption that nodes scoring low on all three measures are unlikely to be critical for transmission, thereby dramatically shrinking the combinatorial space from (\binom{N}{k}) to (\binom{|R₀|}{k}).

In the GA, each individual encodes a set I of exactly k nodes drawn from R₀. The fitness of an individual is defined as the expected number of casualties (i.e., nodes ending in the “Removed” state) under a stochastic SIR simulation. For a given candidate set, the algorithm removes those nodes from the original graph (modeling immunization), randomly selects a single initially infected node, and runs the discrete‑time SIR dynamics with infection probability β = 0.3 and recovery probability γ = 0.3. The simulation proceeds until no infectious nodes remain, and the final number of removed nodes is recorded. Because the initial infected node is chosen at random, the simulation is repeated m = 100 times, and the average casualty count serves as the fitness value. The GA seeks to minimize this expectation.

Selection is performed via tournament selection (size 4). Crossover uses a modified uniform strategy: the two parent chromosomes are concatenated, sorted, and then split into odd‑indexed and even‑indexed halves to form two children, guaranteeing that each child contains k distinct nodes. Mutation occurs with probability 1/k per gene, effectively swapping a gene with another randomly chosen node from R₀. The GA runs for 100 generations with a population size of 100.

The methodology is evaluated on two real‑world networks: (1) the Goodreau Faux Mesa high‑school friendship network (147 nodes, 202 edges) and (2) the US air‑transportation network (500 nodes, 2 980 edges). Two immunization budgets are examined, k = 10 and k = 50, and the candidate pool size l is set to 100 or 200. Baselines include the three single‑metric strategies (selecting the top k nodes by degree, betweenness, or eigenvector) and a random selection baseline.

Results show that on the air‑transportation network the GA consistently yields lower average casualties than any single‑metric strategy, achieving reductions of roughly 12–15 % compared with the best baseline. This improvement is attributed to the network’s heterogeneous hub‑and‑spoke structure, where a node may have modest degree but high betweenness, or vice versa; the GA can combine such complementary nodes in its solution. On the high‑school network, the GA’s performance matches the best single‑metric method (betweenness) and outperforms random selection by about 8 %. The smaller, more homogeneous network makes a single centrality measure already quite effective, yet the GA still offers a modest gain.

The paper’s contributions are threefold: (1) a principled way to prune the search space using multiple centralities, (2) an objective function directly tied to expected epidemic outcomes rather than proxy metrics, and (3) a GA design that preserves diversity and avoids duplicate nodes during crossover. The authors also discuss limitations, such as the use of a static network, a single SIR parameterization, and the absence of cost‑benefit analysis for immunization. Future work is suggested in the directions of dynamic networks, more elaborate compartmental models (SIRS, SEIR), incorporation of heterogeneous vaccine efficacy and deployment costs, and adaptive tuning of the l parameter.

Overall, the study demonstrates that meta‑heuristic optimization, when guided by multi‑metric centrality information, can produce immunization strategies that are competitive with or superior to traditional single‑metric approaches, especially in large, complex networks where the interplay of different notions of node importance is pronounced. This framework offers a practical tool for public‑health planners and infrastructure managers seeking to allocate limited protective resources efficiently.