Attribute analysis of information systems based on elementary soft implications

Soft set theory provides a parameterized treatment of uncertainty, which is closely related to soft computing models like fuzzy sets and rough sets. Based on soft sets and logical formulas over them, this study aims to present a new approach for revealing the causal relationship between values of attributes in an information system. The main procedure of our new method is as follows: First, we choose the attributes to be analyzed and construct some partition soft sets from a given information system. Then we compute the extended union of the obtained partition soft sets, which results in a covering soft set. Further, we transform the obtained covering soft set into a decision soft set and consider logical formulas over it. Next, we calculate various types of soft truth degrees of elementary soft implications. Finally, we can rank attribute values and plot some illustrative graphs, which helps us extract useful knowledge from the given information system. We use several examples, including a classical example given by Pawlak and a practical application concerning IT applying features analysis, to illustrate the newly proposed method and related concepts. In addition, we compare soft attribute analysis with rough attribute analysis and also relate it to soft association rules mining.