Evaluation of Contribution Rate of UAV Swarms System Based on Machine Learning

With the development of unmanned systems, the cross-link coupling between individual drones and swarms has become increasingly strong, making the evaluation of the contribution rate of individual equipment to the entire system more critical. This paper first constructs a performance index system for UAV swarms. Using the entropy weight method, weights are calculated for five performance parameters, calculating and screening out the core indicators: maximum range (27%), payload (23%), and cruising speed (20%). Taking the typical UAV rescue scenario constructed by NetLogo simulation as an example, based on 20 sets of simulation samples (100 experiments in each group), Random Forest(RF), Extreme Learning Machine(ELM), BP neural network(BP-NN) and support vector regression (SVR) models are established to comprehensively predict the contribution rate of UAV and verify its accuracy. Among them, the SVR model demonstrates the best fitting performance. The support vector regression model is optimized by particle swarm optimization (PSO) and quantum particle swarm optimization (QPSO). The kernel width parameter G and penalty parameter C of SVR are optimized globally, and the SVR model optimized by QPSO shows better performance. The experimental results show that the correlation coefficient R2 of the QPSO-SVR model is 0.9984, which is 10.7% higher than that of the traditional SVR, and the root mean square error RMSE is reduced to 0.0015.The results indicate that this method can provide scientific guidance for calculating the system contribution rate in drone swarm rescue mission scenarios. However, the research limitation is that this paper is based on small-scale simulation data, and needs to be extended to actual scenarios or large sample situations in the future.