Cardiovascular disease classification method and device based on multi-branch chain type neural network

Abstract

The invention relates to a cardiovascular disease classification method based on a multi-branch chain type neural network. The multi-branch chain type neural network is provided for solving the problems that a traditional neural network and a machine learning model are poor in classification performance on a small data set, and correlation between labels is ignored. The method comprises the following steps: adding a prediction label sequence into a neural network of each branch so as to utilize the correlation between labels; enabling prediction label sequences added in different branches to adopt different correlation calculation modes to avoid ignoring the diversity of relations between labels. In order to test the performance of the network model, an experiment is carried out on a cardiovascular disease data set provided by a hospital party, and an experiment result shows that compared with other neural networks and machine learning models, the multi-branch chain type neural networkachieves higher performance, and the generalization ability of the model is greatly improved.

Description

technical field [0001] The invention relates to a feedforward neural network, which belongs to the field of deep learning, and in particular to a cardiovascular disease classification method and device based on a multi-branched chain neural network. Background technique [0002] In traditional supervised learning, an example is associated with only one category label, and this type of problem becomes a single-label classification problem. However, in the diagnosis and classification of medical data, a case sample may have multiple different medical record labels at the same time. For example, a patient may have several different diseases at the same time. The present invention calls this type of problem a multi-label classification problem. [0003] The problem of multi-label classification has gradually become an emerging hot research point in the field of machine learning. In multi-label classification, the most common method is the problem conversion method, that is, b...

AI Technical Summary

Problems solved by technology

[0010] As mentioned above, there are obvious defects in the problem transformation method, and the methods proposed by the predecessors to deal with the relationship between labels are often based on a single classification model; when a label is classified incorrectly in the classification, the error will be Seriously affect the classification results of subsequent series of labels; therefore, in multi-label classification, not only the classification order of labels is important, but also the performance of the classifier is very critical, while the performance of a single classification model is very limited

Images