Confidence Assessment Method for Association Rule Mining Based on Task Characteristics

With the development of big data technology, tracking and monitoring specific objectives often require the mutual corroboration and supplementation of data from multiple sources. However, due to constraints such as geographical environment, monitoring equipment, and target attributes, some source information may be inadequately sampled and incomplete, leading to data incompleteness. This incompleteness results in significant disparities between the computed results in real operational scenarios and theoretical expectations. In response to the variation of credibility of multisource data mining analysis results caused by incomplete data, this paper proposes a confidence assessment method for association rule mining based on task characteristics, which introduces a human-in-the-loop contribution assessment process based on the traditional Dempster-Shafer (D-S) evidence theory. This process combines and quantifies confidence in the results of multi-source evidence mining and analysis. Through simulation experiments, the algorithm has been shown to effectively adapt to the mining and classification in situations of incomplete data. The algorithm significantly improves adaptability across various scenarios while ensuring timeliness.