Fundamentals of forecasting indicators of economic activity of executive bodies for sustainable development

Chupina, Zhanna; Sokolnikova, Olga; Yurchenko, Oksana; Ryabinina, E. V.; Veselko, Alena A.

doi:10.1051/e3sconf/202124412026

Cited by 1 publication

(3 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The maximum value is obtained by using the random gradient ascending method, and the optimal parameters and the partial derivatives are obtained. In the RBM, the final parameter results are obtained by using the K-step contrast divergence algorithm, as shown in equations ( 9) (10) and (11). w=w+η∆w ( 9) c=c+η∆c ( 10) b=b+η∆b (11) In the above equations, η is the learning rate, w, c, b are the adjustment parameters.…”

Section: Classification and Partition Of The Rbm Algorithmmentioning

confidence: 99%

“…In the RBM, the final parameter results are obtained by using the K-step contrast divergence algorithm, as shown in equations ( 9) (10) and (11). w=w+η∆w ( 9) c=c+η∆c ( 10) b=b+η∆b (11) In the above equations, η is the learning rate, w, c, b are the adjustment parameters.…”

Section: Classification and Partition Of The Rbm Algorithmmentioning

confidence: 99%

“…Once there occurs debts, domestic enterprises will bear huge losses [10]. Therefore, the study on export credit risks is very important, and the credit risk assessment system has far-reaching significance for the sustainable development of the regional economy [11].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Application of Deep Belief Network for Sustainable Development via Deep Learning to Export Credit Risk Assessment Under the Belt and Road Strategy

Li,

Pu,

Song

et al. 2023

Advances in Transdisciplinary Engineering

View full text Add to dashboard Cite

The purpose of the study is to reduce the export credit risk of enterprises, and the export credit risk assessment of enterprises under the belt and road strategy based on deep learning is discussed. First, the research background and deep belief networks are introduced. Second, the contrast divergence algorithm based on the deep belief model is improved on the restricted Boltzmann machine. Finally, the deep belief network of classification and partition is constructed and simulated. The results show that the test accuracy of the classification and partition of the restricted Boltzmann machine (CPRBM) constructed is higher than that of the restricted Boltzmann machine (RBM). When the accuracy of the algorithm is verified under the condition of unbalanced two classification samples in a relatively small amount of datasets, the accuracy of the CPRBM algorithm is 93.71%, and the accuracy of the RBM algorithm is 89.86%. In the optimization stage of the deep belief networks, the convergence rate of the CPRBM is slower than that of the RBM. Since the optimized system increases the penalty term in the first training stage of the deep belief networks, the penalty term is canceled in the second stage of optimization. At three time points, the algorithm accuracy of the CPRBM is higher than that of the RBM. The simulation results are consistent with the previous experimental results. Although the accuracy is not high at the third time point, the CPRBM algorithm still has some advantages. Compared with the accuracy of the support vector machine (SVM) and the deep extreme learning machine (DELM), the CPRBM algorithm based on deep belief networks has the highest accuracy at any time point. The CPRBM algorithm constructed has obvious advantages compared with common models, and the overall performance of the algorithm is better. The conclusions provide the support for the sustainable development of the economy under the Belt and Road strategy.

show abstract

Section: Classification and Partition Of The Rbm Algorithmmentioning

confidence: 99%