Error-Dependency Relationships for the Naïve Bayes Classifier with Binary Features

• Published in: IEEE transactions on pattern analysis and machine intelligence Vol. 30; no. 4; pp. 735 - 740
• Main Authors: Kuncheva, L.I., Hoare, Z.S.J.
• Format: Journal Article
• Language: English
• Published: Los Alamitos, CA IEEE 01.04.2008; IEEE Computer Society; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Applied sciences; Artificial Intelligence; Bayes Theorem; Classifier design and evaluation; Computer science; control theory; systems; Computer Simulation; Data Interpretation, Statistical; Dependency; Error analysis; Exact sciences and technology; Feature evaluation and selection; Logistic Models; Naive Bayes; Niobium; Pattern Recognition; Pattern Recognition, Automated - methods; Probability density function; Q measurement; Reproducibility of Results; Robustness; Sensitivity and Specificity; Signal Processing, Computer-Assisted; Statistics; Pattern Recognition; Feature evaluation and selection; Naive Bayes; Dependency; Classifier design and evaluation; Bayes estimation; Naïve Bayes classifier; Statistical analysis; Probabilistic approach; Computer simulation; Classification; dependency between features; Pattern recognition; Q statistic; Pattern analysis; Artificial intelligence; classification error
• ISSN: 0162-8828; 1939-3539
• DOI: 10.1109/TPAMI.2007.70845

We derive a tight dependency-related bound on the difference between the NB error and Bayes error for the case of two binary features and two classes. A measure of feature dependency is proposed for multiple features. Simulations and experiments with 23 real data sets were carried out.