Deep learning-based overseas mining investment risk evaluation method

Abstract

The invention discloses an overseas mining investment risk evaluation method based on deep learning. The method comprises the following steps of preparing original data; designing a WGAN-GP model to carry out data enhancement on the original data; performing PCA (principal component analysis) dimension reduction on the risk evaluation indexes; and constructing an overseas mining investment risk evaluation model based on the DNN deep neural network. According to the method, data enhancement is realized, a deep neural network model is constructed, and the overseas investment target mining country investment risk is evaluated.

Description

technical field [0001] The present invention relates to the field of artificial intelligence for overseas investment risk assessment, and more specifically, relates to a deep learning-based overseas mining investment risk assessment method. Background technique [0002] Commonly used methods for quantitative evaluation of risk include fuzzy theory, analytic hierarchy process and gray theory, as well as entropy weight and BP neural network method introduced in recent years. The application of these methods has certain limitations. [0003] Fuzzy theory, on the basis of determining the evaluation factors, factor evaluation grade division standards and weights, using the principle of fuzzy set transformation, using the degree of membership to describe each factor and the fuzzy boundary of the factor, constructing the evaluation matrix, and through multi-layer composite operations, finally Determine the evaluation object level. [0004] Analytic Hierarchy Process divides the in...

AI Technical Summary

Problems solved by technology

[0008] (1) Mainly rely on artificial design features, inevitably mixed with subjective factors of researchers

[0009] (2) The traditional linear method also has an overly strong "linear" setting, and the influence of various factors on risk is often nonlinear

Even for more effective methods such as BP neural network, in addition to problems such as overfitting and gradient disappearance, because it is still a shallow learning method, the processing ability of input features is limited, and the generalization ability is restricted when solving complex classification problems.

[0010] (3) The training of the deep neural network model often requires a large amount of data. It is difficult to obtain overseas mining risk assessment data, and there is a problem of overfitting with too little data.

[0011] (4) Neural network parameters are redundant, input parameters are poorly correlated, and the network model is complex

However, the reminder and evaluation of investment risk is essentially a high-dimensional classification with high complexity.

Images