An efficient uncertainty measure with dynamic update mechanisms

With the extensive adoption of information technology, the data we encounter today is frequently dynamic and subject to change over time. To facilitate timely decision-making, it is crucial to possess a measure that can swiftly identify and continuously update the inherent uncertainty present in the data. In this paper, we present a measure of weighted uncertainty, referred to as WCE, and investigate methods for its dynamic updating within information systems. Initially, the granularity of the universe is established based on binary relations derived from each attribute, which is subsequently utilized to assign weights. Following this, we employ conditional entropy to assess the uncertainty level of the target concept concerning each attribute. Ultimately, the overall uncertainty of the information system is computed by weighting the uncertainty associated with each attribute. To enhance the intuitiveness and simplicity of dynamic updates for weighted uncertainty more intuitive and straightforward, we transform the WCE into matrix form. We then delve into the dynamic updating mechanism, examining how the core matrices are modified in response to variations in data volume or attributes. Finally, numerical experiments conducted on UCI datasets demonstrate that the proposed WCE measure is responsive to diverse data changes. Its updating approach for dynamic information systems can significantly reduce time consumption.