A data classification storage method and device and a storage medium

Abstract

The application discloses a data classification storage method and device and a storage medium, relates to the technical field of computer data storage, and comprises the following steps: acquiring multi-modal data; inputting the multi-modal data into an intelligent classification model to obtain a classification result; wherein the intelligent classification model is obtained through a bootstrap training strategy, and the bootstrap training strategy comprises the following steps: generating weak labels for unlabeled data based on a labeling function generated based on a preset rule, and performing cold start training on an initial classification model; screening data with prediction uncertainty satisfying a preset condition based on an active learning mechanism to perform labeling backfilling, iteratively updating the initial classification model, and obtaining the intelligent classification model; and based on the classification result, determining a target storage node of the multi-modal data in a distributed storage architecture, and writing the multi-modal data into the target storage node. The model cold start problem is solved through the bootstrap training strategy, and massive manual labeling is not required. In addition, the active learning and labeling backfilling are used, so that the classification accuracy of the multi-modal data is improved.

Images