A novel data credibility-centric multi-model-based complex systems modeling approach for UAV capability evaluation

Chenhao Yu, Junyi Zhu, Lei-Lei Chang, You Cao, Xiaobin Xu, Zhiyong Hao

Abstract: Complex systems modeling has wide applications in theory and practice. A new approach is proposed by recognizing data credibility (DC) using multiple machine learning (ML) approaches, named DCML. There are two major components and theoretical contributions of the proposed approach: first, identifying less-credible data with a single ML approach, and second, cross-identifying these less-credible data with multiple ML approaches. A practical case of capability evaluation of the Unmanned Aerial Vehicle (UAV) is studied for validating the effectiveness of DCML. Case study results show that (1) The proposed DCML approach demonstrates a proficient ability to identify less credible data, (2) The validations with various ML methods prove effective, but the efficacy of the method is not necessarily proportional to the number of methods employed, (3) The combination of backpropagation neural network (BPNN) and gaussian process regression (GPR) yields the most favorable outcomes.