Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CNC-TP: Classifier Nominal Concept Based on Top-Pertinent Attributes

Yasmine Souissi, Fabrice Boissier, Nida Meddouri

TL;DR

This work advances FCA-based classification by introducing CNC-TP, which selects top-pertinent attributes via the Gain Ratio and constructs concepts and rules from a reduced, more informative feature subset. It presents two concept extraction strategies (All Values vs Relevant Values) and evaluates them across 16 UCI datasets using a robust 10-fold cross-validation protocol, achieving strong accuracy and AUC metrics for the All-Values variant at $p\approx0.775$. CNC-TP demonstrates competitive performance against rule-based, tree-based, Bayesian, and statistical models while preserving interpretability through its rule-based, lattice-grounded structure and a rejection mechanism for uncertain cases. The results underscore the practical value of adaptive feature selection in FCA-based classifiers and point to future work integrating hybrid and fuzzy approaches to further enhance robustness and applicability.

Abstract

Knowledge Discovery in Databases (KDD) aims to exploit the vast amounts of data generated daily across various domains of computer applications. Its objective is to extract hidden and meaningful knowledge from datasets through a structured process comprising several key steps: data selection, preprocessing, transformation, data mining, and visualization. Among the core data mining techniques are classification and clustering. Classification involves predicting the class of new instances using a classifier trained on labeled data. Several approaches have been proposed in the literature, including Decision Tree Induction, Bayesian classifiers, Nearest Neighbor search, Neural Networks, Support Vector Machines, and Formal Concept Analysis (FCA). The last one is recognized as an effective approach for interpretable and explainable learning. It is grounded in the mathematical structure of the concept lattice, which enables the generation of formal concepts and the discovery of hidden relationships among them. In this paper, we present a state-of-theart review of FCA-based classifiers. We explore various methods for computing closure operators from nominal data and introduce a novel approach for constructing a partial concept lattice that focuses on the most relevant concepts. Experimental results are provided to demonstrate the efficiency of the proposed method.

CNC-TP: Classifier Nominal Concept Based on Top-Pertinent Attributes

TL;DR

This work advances FCA-based classification by introducing CNC-TP, which selects top-pertinent attributes via the Gain Ratio and constructs concepts and rules from a reduced, more informative feature subset. It presents two concept extraction strategies (All Values vs Relevant Values) and evaluates them across 16 UCI datasets using a robust 10-fold cross-validation protocol, achieving strong accuracy and AUC metrics for the All-Values variant at . CNC-TP demonstrates competitive performance against rule-based, tree-based, Bayesian, and statistical models while preserving interpretability through its rule-based, lattice-grounded structure and a rejection mechanism for uncertain cases. The results underscore the practical value of adaptive feature selection in FCA-based classifiers and point to future work integrating hybrid and fuzzy approaches to further enhance robustness and applicability.

Abstract

Knowledge Discovery in Databases (KDD) aims to exploit the vast amounts of data generated daily across various domains of computer applications. Its objective is to extract hidden and meaningful knowledge from datasets through a structured process comprising several key steps: data selection, preprocessing, transformation, data mining, and visualization. Among the core data mining techniques are classification and clustering. Classification involves predicting the class of new instances using a classifier trained on labeled data. Several approaches have been proposed in the literature, including Decision Tree Induction, Bayesian classifiers, Nearest Neighbor search, Neural Networks, Support Vector Machines, and Formal Concept Analysis (FCA). The last one is recognized as an effective approach for interpretable and explainable learning. It is grounded in the mathematical structure of the concept lattice, which enables the generation of formal concepts and the discovery of hidden relationships among them. In this paper, we present a state-of-theart review of FCA-based classifiers. We explore various methods for computing closure operators from nominal data and introduce a novel approach for constructing a partial concept lattice that focuses on the most relevant concepts. Experimental results are provided to demonstrate the efficiency of the proposed method.
Paper Structure (15 sections, 3 equations, 13 figures, 4 tables, 1 algorithm)

This paper contains 15 sections, 3 equations, 13 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Formal Concept Analysis
  3. State of the Art for Classification Based on Formal Concept Analysis
  4. Proposed Approach: CNC-TP
  5. Classifier Construction
  6. Concept Extraction Strategies
  7. Concepts and Rules Generation
  8. Classification by CNC-TP
  9. Experimentation
  10. Experimental Protocol
  11. Data Sets
  12. CNC-TP-AV: Top Pertinent Attributes with All Values
  13. CNC-TP-RV: Top Pertinent Attributes with Relevant Values
  14. CNC-TP Compared to Other Classifiers
  15. Conclusion

Figures (13)

  • Figure 1: Concept lattice generated from the previous formal context using Galicia
  • Figure 2: CNC-TP with all values: evolution of correct, incorrect, and unclassified instance rates as a function of the threshold
  • Figure 3: CNC-TP with all values: evolution of additional performance metrics as a function of the threshold
  • Figure 4: CNC-TP with relevant values: evolution of correct, incorrect, and unclassified instance rates
  • Figure 5: CNC-TP with relevant values: evolution of additional performance metrics
  • ...and 8 more figures