Uncertainty measurement for heterogeneous data: an application in attribute reduction

In the era of big data, multimedia, hyper-media and social networks are emerging, and the amount of information is growing rapidly. When people participate in the process of massive data processing, they will encounter data with different structures, so data has heterogeneity. How to acquire hidden and valuable knowledge from heterogeneous data and measure its uncertainty is an important problem in artificial intelligence. This paper investigates uncertainty measurement for heterogeneous data and gives its application in attribute reduction. The concept of a heterogeneous information system (HIS) is first proposed. Then, an equivalence relation on the object set is constructed. Next, uncertainty measurement for a HIS is investigated, a numerical experiment is given, and dispersion analysis, correlation analysis, and Friedman test and Bonferroni-Dunn test in statistics are conducted. Finally, as an application of the proposed measures, attribute reduction in a HIS is studied, and the corresponding algorithms and their analysis are proposed.

Automated summary

This paper investigates uncertainty measurement for heterogeneous data and its application in attribute reduction. It first proposes the concept of a heterogeneous information system (HIS) and constructs an equivalence relation on the object set. The paper then conducts numerical experiments and statistical analysis, studying attribute reduction in a HIS and corresponding algorithms.