Hepatic encephalopathy prediction model of multi-scale neural network based on data enhancement and modeling method

The invention discloses a hepatic encephalopathy prediction model of a multi-scale neural network based on data enhancement and a modeling method. The prediction model comprises a feature selection module, a data enhancement module, a deep convolution module and a multi-scale convolution module. And the feature selection module comprises feature vectors f1 to fn, a feature vector f (childpuff), a feature vector f (childpuff), a feature vector f (MELD) and a feature vector f (MELDgrade). The feature selection module comprises the feature vectors f1 to fn, the feature vector f (childpuff), the feature vector f (MELD) and the feature vector f (MELDgrade). And the data enhancement module comprises data X11 to X1n, data X21 to X2n, data X1i, data X1j, data XSi, data X2i and data X2j. The deep convolution module comprises data Xi and a data set XS. The multi-scale convolution module comprises a data set XS. According to the method, the self-attention mechanism is used for paying attention to more important features