Relational Approach to the L-Fuzzy Concept Analysis

Modern industrial production systems benefit from the classification and processing of objects and their attributes. In general, the object classification procedure can coincide with vagueness. Vagueness is a common problem in object analysis that exists at various stages of classification, including ambiguity in input data, overlapping boundaries between classes or regions, and uncertainty in defining or extracting the properties and relationships of objects. To manage the ambiguity mentioned in the classification of objects, using a framework for L-fuzzy relations, and displaying such uncertainties by it can be a solution. Obtaining the least unreliable and uncertain output associated with the original data is the main concern of this thesis. Therefore, my general approach to this research can be categorized as follows: We developed an L-Fuzzy Concept Analysis as a generalization of a regular Concept Analysis. We start our work by providing the input data. Data is stored in a table (database). The next step is the creation of the contexts and concepts from the given original data using some structures. In the next stage, rules, or patterns (Attribute Implications) from the data will be generated. This includes all rules and a minimal base of rules. All of them are using L-fuzziness due to uncertainty. This requires L-fuzzy relations that will be implemented as L -valued matrices. In the end, everything is nicely packed in a convenient application and implemented in Java programming language. Generally, our approach is done in an algebraic framework that covers both regular and L -Fuzzy FCA, simultaneously. The tables we started with are already L-valued (not crisp) in our implementation. In other words, we work with the L-Fuzzy data directly. This is the idea here. We start with vague data. In simple terms, the data is shown using L -valued tables (vague data) trying to relate objects with their attributes at the start of the implementation. Generating attribute implications from many-valued contexts by a relational theory is the purpose of this thesis, i.e, a range of degrees is used to indicate the relationship between objects and their properties. The smallest degree corresponds to the classical no and the greatest degree corresponds to the classical yes in the table