An improved limited and variable precision rough set model for rule acquisition based on pansystems methodology

The purpose of this paper is to acquire doubly variable precision‐based knowledge rules from incomplete decision tables (IDTs) in the framework of pansystems methodology. It suggests a new variable precision limited tolerance – a special pansystems relation – rough set model with precision inclusion and a reduct procedure in which it overcomes the non‐monotony in forming tolerance classes when reducing an attribute from attribute set.

Through introducing variable precision and limited tolerance relation in IDT, it constructs symmetric binary relation, dissimilar to non‐symmetric relation proposed by others, and then forms tolerance classes. It proposes a new reduction procedure with absolute value calculation to avoid tolerance classes being non‐monotone. Using variable inclusion, it obtains lower and upper approximations with noises.

Tolerance classes are not monotone with the reduction of attribute from attribute set in the proposed variable precision and limited tolerance relation, but it remains symmetry. Proposed reduction procedure with absolute value calculation is a new approach in adjudging whether a reduct equals to the original whole attribute set within a error range or not.

Using variable precision and limited rough set model with variable inclusion to mine deep knowledge from IDT is a paradise in knowledge discovery in dealing with non‐determinative and vague problems.

The formation of symmetric tolerance relation is natural. The reduction procedure with absolute value calculation is new and not similar to those existed in literatures.