Multivariate time series data filling method based on real data enhancement

Abstract

The invention discloses a multivariate time series data missing value filling method based on real data enhancement. The method comprises the steps of compressing real data into a data representationvector through an encoder, then generating a complete data vector from the data representation vector through a generator in combination with real data enhancement, and finally distinguishing the realdata and the generated data through a discriminator. According to the whole model, the loss of an encoder, a generator and a discriminator is optimized through generative adversarial training, so that data generated by the generator are close enough to real data to confuse the discriminator, and finally, missing values in multivariate time series data are filled with the generated data of the generator. According to the method, a generative adversarial network with an encoder is taken as a framework, and real data are applied to the encoder, the generator and the discriminator of the model, so that the data generated by the model are more real and reliable, and can be effectively used for filling missing values.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence, and in particular relates to a multivariate time-series data filling method based on real data enhancement. Background technique [0002] Multivariate time-series data commonly exists in various practical scenarios in the real world, such as electronic medical records regularly filed by hospitals, stock prices that change daily in the stock market, and climate factors constantly monitored by the Meteorological Bureau, etc. These data are recorded at multiple moments, and the records at each moment contain multivariate time-series data of multiple elements, which fully preserves the overall change law of the data in the corresponding scene. Analyzing the multivariate time-series data of the corresponding scene can conduct a comprehensive analysis of the scene problems and make predictions on the development trend of things. [0003] Due to irregular data collection methods, unstab...

AI Technical Summary

Problems solved by technology

However, none of these methods consider the introduction of real data in the generator, which limits the performance of generative models.

During the data generation process of the generator, the input of each step is generated by the generator itself and does not receive other inputs, which will cause wrong input to continue to affect the generation of subsequent data

Images