Research on Cluster Innovation Network Optimization Based on Tabu Search Algorithm—A Case Study of Northeast New Energy Automobile Industry Cluster

Taking the new energy vehicle (NEV) industry in northeast China as the research object, this paper constructs the cluster innovation network of the northeast NEV industry by the snowball sampling method. Based on the tabu search algorithm, three local optimization models are constructed under the strategies of edge adjustment (rewiring, Type-I), edge reinforcement (addition, Type-II), and edge reduction (deletion, Type-III), and the network optimization processes under different types are empirically analyzed from the perspectives of network structure differentiation and the degree of optimization goal realization. The results show that: (i) all three types of local optimization networks exhibit only modest changes in basic structural metrics relative to the actual network, and the changes in degree sequence and degree distribution on the actual network structure are not very significant; (ii) the three local optimization networks adjust the network structure mainly by weakening the connections of the top core-tier nodes while strengthening high-degree intermediate-tier nodes together with a long tail of edge-level nodes (Type-I), strengthening the connections among intermediate-tier nodes together with selected peripheral and core-tier nodes (Type-II), and reducing the redundant connections of several intermediate-tier and peripheral nodes (Type-III); (iii) in addition, the Type-II (edge-addition) local optimization network achieves the best performance on both network stability and aggregate innovation benefit, and its first rank is robust to equal-adjustment budgets, to tie-creation costs of up to four times the per-tie maintenance cost, and to every alternative diffusion dynamic and to intensity-weighted evaluation tested, whereas the ranking of the remaining two strategies is specification-dependent (Type-I is a cost-effective second-best under the baseline specification, without expanding the edge count); the locally optimized networks can effectively realize and enhance the network function, strengthen the stability of the northeast NEV cluster innovation network, and further improve the social benefits of the NEV industry. The baseline network comprises 135 sampled organizations spanning large state-owned OEMs, medium-sized suppliers, and small specialized firms. Limitations include the single-case, cross-sectional design and the geographic concentration of the seed set.