Abstract:The rough set theory provides a powerful approach for attributes reduction and data analysis. The variable precision rough set (VPRS) model, an extension of the original rough set approach, tolerates misclassifications of the training data to some degree, which promotes the applications of rough set theory in inconsistent information systems. However, in most existing algorithms of feature reduction based on VPRS, the precision parameter (β) is introduced as prior knowledge, which restricts their applications because it is not clear how to set the β value. By studying β-consistency in the measurement of a decision table and the threshold value of the β-consistent decision table, this paper presents an algorithm for automatic determination of the precision parameter value from a decision table based on VPRS. At the same time, the precision parameter value from our proposed method is compared with the thresholds from the decision-theoretic rough set (DTRS). The influences of the precision parameter are also discussed on attribute reduction, which shows the necessity of the estimated precision parameter from a decision table. The simulation results including VPRS and other classification methods in real data further indicate that different precision parameter values make a great difference on rules and setting a precise parameter near the threshold value of the β-consistent decision table can precisely reflect the decision distribution of the decision table.

Probabilistic Approximation Under Incomplete Information Systems

Imprecise Probabilities From Imprecise Descriptions Of Real Numbers

Theory and Algorithm Based on the General Similar Relationship Between the Approximate Reduction

Approximation Space Based on the Probability of Default Rules Mining Algorithm

VPRS-Based Knowledge Discovery Approach in Incomplete Information System

Fuzzy Probabilistic Approximation Spaces and Their Information Measures

Knowledge Approximation and Rule Acquisition Based on VPRS in Ordered Information Systems

Imputation for categorical attributes with probabilistic reasoning

Probabilistic Safety Regions Via Finite Families of Scalable Classifiers

The Model of Bidirectional Ps-Rough Sets in Probabilistic Approximation Space and Its Application

Rough Set Model and Precision Reduction in Incomplete and Fuzzy Decision Information Systems

Approximate Accuracy Approaches to Attribute Reduction for Information Systems

Enhancement of Approximation Spaces by the Use of Primals and Neighborhood

Entropy Based Attribute Reduction Approach For Incomplete Decision Table

A Novel Approach of Rough Conditional Entropy-Based Attribute Selection for Incomplete Decision System

Automatic determination about precision parameter value based on inclusion degree with variable precision rough set model

An Approximate Reduction Algorithm Based On Conditional Entropy

Rough Approximation In Fuzzy Decision Systems

Neighborhood Operator Systems and Approximations

A new data filling approach based on probability analysis in incomplete soft sets

Approximate Information States for Worst-Case Control and Learning in Uncertain Systems