Online early warning method for big data exception of medical cloud platform based on statistical generation model

Abstract

The invention discloses an online early warning method for big data exception of a medical cloud platform based on a statistical generation model. Through a two-step filtering method including affinetransformation and direction smooth filtering, time sequence fragment data is filtered, so that similar points in the time sequence fragment data are removed, a small number of feature points are reserved, the analysis data volume is reduced, and meanwhile, a data basis is provided for a statistical generation model. For searching of abnormal early warning samples, the method adopts an online Gaussian mixture statistical generation model, the model fits probability distribution of a full life cycle of medical data, the occurrence probability of real-time time sequence samples can be calculated, and a low probability sequence in the real-time time sequence samples is selected as an early warning sample, so that online early warning of big data abnormality of a medical cloud platform is realized.

Description

technical field [0001] The invention relates to a big data abnormality judgment and early warning method of a medical cloud platform, in particular to a big data abnormality judgment and early warning method of a medical cloud platform based on a statistical generation model. Background technique [0002] A large amount of drug manufacturing, storage and circulation data, as well as data on patients' medication habits and methods are stored in the medical cloud platform. These data can often reflect the temporal and spatial distribution characteristics and future development trends of various drugs and related diseases. Industry workers may Concerned about the changes in the spatio-temporal distribution of a certain type of drug or a brand of drugs, or looking for potential causal relationships among all changes. In the face of massive big data, relying on regular reports in the past cannot meet the needs of the industry in terms of timeliness and operability, so it needs to...

AI Technical Summary

Problems solved by technology

Due to the relatively smooth changes in data such as manufacturing, logistics, and regional circulation in the medical field, the feature points extracted by the current method are still too dense, and a large number of similar and repeated features are retained, so that feature extraction cannot improve the efficiency of algorithm execution; The method of dynamic time window or clustering depends on the rationality of the definition of distance measurement for the sample sequence. For the data of medical cloud platform, there is no ideal distance measurement method at present.

Images