Semi-supervised classification algorithm for data streams based on real-time monitoring and early warning

With the continuous advancement of geological monitoring technology, real-time monitoring and early warning systems are playing an increasingly significant role in preventing geological disasters. However, facing massive, high-speed, and dynamically changing geological data streams, how to classify and issue warnings efficiently and accurately has become an urgent issue to be addressed. This paper proposes the SSCMW algorithm, which aims to effectively mine and issue warnings for geological data streams using a small amount of labeled data and a large amount of unlabeled data. To better address different types of concept drift, SSCMW sets up two classifier pools. The classifiers in the first pool undergo incremental updates and component replacements. The classifiers in the second pool are replicated from the first pool, allowing for better handling of recurring concepts. Furthermore, we calculate the deviation of classifiers based on the local structure of clustering to detect recurring concepts. We compared this algorithm with existing baseline methods on multiple datasets, and the experimental results confirmed the effectiveness of the algorithm.