科研之友
微信
新浪微博
Facebook
分享链接
摘要
Aiming at defects of slow learning speed, trapped in local solution and inaccurate operating result of traditional BP neural network in the practical application of drought prediction, a drought predic- tion model based on parallel ensemble learning algorithm of good point set glowworm swarm optimization algorithm (GPSGSO) and back propagation neural network (BPNN) is proposed. Firstly, a new kind of improved glowworm swarm algorithm based on good point set theory and inertia weight function is constructed, and the validity of the algorithm is analyzed theoretically. Secondly, GPSGSO algorithm and BPNN are combined to construct parallel ensemble learning algorithm. GPSGSO is used to optimize the weight and threshold of BPNN, and the ensemble strategy is carried out for the best weights and thresholds. Finally, the parallel ensemble learning algorithm is applied to the prediction of agricultural drought disaster, which can accurately determine the drought level. The effectiveness of the GPSGSO algorithm in terms of convergence speed, accuracy and stability is verified by 8 Benchmark functions. At the same time, agricultural meteorological data of Northern Anhui Province is used to simulate validate experiment, the experimental results show that the algorithm has obvious advantages over the traditional BPNN, GSO-BPNN and GA-BPNN algorithm in terms of convergence speed, operation accuracy and sta- bility. Therefore, the drought prediction model based on GPSGSO-BPNN parallel learning algorithm can effectively improve the accuracy of agricultural drought prediction.
