Financial unbalanced data classification method based on deep learning

The invention relates to a financial unbalanced data classification method based on deep learning, and belongs to the technical field of data mining. The input layer takes financial digital data and text data as input of a classification model; the preprocessing layer preprocesses the data of the input layer; the text feature extraction layer performs word embedding on text data output by the preprocessing layer by using a BERT pre-training model, extracts sequence features of a long text by using a BiGRU bidirectional structure, and extracts local features and important global information of the text by using a convolutional neural network fused with a multi-head self-attention mechanism; the global feature interaction layer is used for splicing the digital data output by the preprocessing layer and the output of the text feature extraction layer, and mining associated features between input features and classification results by using a residual cross network; and the decision-making layer carries out batch