Method and System for Oversampling Imbalanced Classification Time-Series Data

Abstract

The present invention relates to the technical field of data processing, and specifically to a method and system for oversampling imbalanced classification time-series data. The method includes: mapping the feature representation extracted from imbalanced classification time-series data from a high-dimensional space to a low-dimensional space, and introducing a supervised contrastive mechanism to adjust the distribution of the time- series data in the low-dimensional space; identifying high-quality samples among minority class samples according to the distribution density of the time-series data in the low- dimensional space; generating new samples through interpolation based on the high- quality samples; verifying the new samples, and retaining the samples that pass the verification to obtain the time-series data after oversampling processing. Compared with the prior art, the present invention improves the quality and balance of time-series data during oversampling.