A weight optimization method for chemical safety evaluation indicators based on the bipartite graph and random walk

In the chemical safety evaluation system, the optimization of indicator weights needs to take both experts' evaluations and the feedback on accident influences into account. Thus, this paper proposes a comprehensive weighting method based on the association bipartite graph (ABG). The accident influences and correlation intensity between the accident and the evaluation indicators are calculated on the ABG. A random walk algorithm, which integrates the objective influences of the accidents and the subjective evaluations of experts, is designed to realize the weight optimization. Experiments prove the effectiveness of the proposed method from the perspectives of weight ranking and fitting degree.