Dynamic medical data oriented causal feature extraction method

Abstract

The invention discloses a dynamic medical data oriented causal feature extraction method, which comprises the following steps: 1, extracting features of a medical data set, and constructing / updating a feature set of the medical data; 2, according to the extracted features and the current medical data, summarizing the medical data information in the form of statistical information by constructing / updating a full-dimensional tree structure; 3, assigning an initial feature subset of the causal feature extraction by using an existing key feature (factor) set, namely initializing an optimal subset; 4, selecting an optimal feature subset by utilizing a causal relationship inference theory and algorithm through iteration of addition and deletion operations, namely obtaining key feature information of the current medical data; and 5, when new medical data arrives, repeating the steps 1-4, and selecting real-time key features from the dynamic medical data. According to the method, the key features can be found from the dynamic medical data, and the extraction of the key features is more interpretable and robust.

Description

technical field [0001] The invention belongs to the field of data mining, relates to technologies such as causal discovery and artificial intelligence, and specifically relates to a causal feature extraction method for dynamic medical data. Background technique [0002] With the rapid development of big data technology, the use of information technology to process health and medical big data to protect the health of residents has become one of the hottest focuses. Among them, health care big data refers to data related to health care generated in the process of people's disease prevention and health management. However, most of the medical data are dynamically generated, huge and messy, and medical staff cannot obtain key information from numerous items in a timely and accurate manner. Therefore, while capturing and summarizing dynamic medical data, it is of great significance how to extract key feature information in current medical data in real time in order to better ass...

AI Technical Summary

Problems solved by technology

Correlations are designed to capture co-occurrences between features and categorical variables, so features selected by correlations do not provide convincing explanations for predictive models

[0006] (2) Lack of robustness

The prediction model constructed by the feature subset selected by the correlation between features and categorical variables, when applied to other data sets with the same distribution, its performance may be significantly reduced, and the effect will be much worse than that of the verification, that is, the model Unreliable

Images