Online Change Point Detection in Open Feature Spaces

In the real world, data is often vast and subject to rapid change. Consequently, an increasing number of researchers have turned their attention to data streaming learning to mine the features present in the data in recent years. Nevertheless, characterizing data in such scenarios frequently presents challenges in the form of missing values or outliers, stemming from sensor wear and tear or inadequate information collection. These issues are prevalent and inescapable in real-world applications. In this paper, we propose the FESL with online imputation algorithm for mixed data using the Gaussian copula and also introduce change point detection to detect outliers. Specifically, our methodology employs the Gaussian copula algorithm as the foundation for imputing missing values, effectively restoring missing data points by modeling the underlying correlations within the dataset. This approach significantly contributes to enhancing data completeness and coherence. Additionally, our framework introduces a change point detection technique designed to identify abrupt transitions within the data stream, thereby capturing shifts and anomalies in the data distribution. Through experimental results, we provide evidence of the efficacy of our approach when addressing similar challenges.