Correlation analysis based on neural network copula function

The joint-distribution function between variables plays an important role in reliability analysis. A method is proposed for constructing the function using a neural network, which is used to construct a copula model under arbitrarily measured data, including the input and output values of the neural network using an empirical cumulative distribution. Three traditional copula function models are constructed based on the Kendall rank-correlation coefficients. Based on the Euclidean distance method, the neural network copula and three copula function models are compared.