Multi-Interval Discretization of Continuous-Valued Attributes for Classification Learning

By Usama M. Fayyad, K. Irani. 1993

Subfields: Evaluation & Benchmarks, Statistical Learning Theory

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Evaluation & Benchmarks

1. Francisco Herrera
2. Salvador Garcia
3. Huan Liu
4. Lei Yu
5. G. Holmes
6. Yew-Soon Ong
7. Eibe Frank
8. M. J. Jesús

In Statistical Learning Theory

1. Cynthia Rudin
2. S. Kotsiantis
3. Houtao Deng
4. Huan Liu
5. Lei Yu
6. Foster J. Provost
7. Jian Pei
8. B. Liu

Citations in the corpus

• Black-Box vs. White-Box: Understanding Their Advantages and Weaknesses From a Practical Point of View (2019, background): “discretization for all numerical features [131], and rule mining strategies provides a large number of rules, which”
• MurTree: Optimal Decision Trees via Dynamic Programming and Search (2020, methodology): “…into binary datasets as a preprocessing step using a supervised discretisation algorithm based on the minimum description length principle (MDLP) by Fayyad and Irani (1993), effectively converting each feature into a categorical feature based on the statistical significance of the feature values for…”
• A critical review of multi-objective optimization in data mining: a position paper (2004, methodology): “This principle is often used to favour the discovery of knowledge that is both accurate and simple [Tuzhilin 2002], [Quinlan & Rivest 1989], [Fayyad & Irani 1993].”
• An Empirical Comparison of Voting Classification Algorithms: Bagging, Boosting, and Variants (1999, methodology): “Quinlan (1996) wrote that his better results using AdaBoost compared with Freund & Schapire (1996) may be due to his use of reweighting compared with their use of resampling.”
• Toward integrating feature selection algorithms for classification and clustering (2005, background): “discretization [28], [51] and missing value treatment, while”
• Automated breast cancer detection and classification using ultrasound images: A survey (2015, methodology): “Other feature extraction techniques such as factor analysis (FA) [162], independent component analysis (ICA) [158], and discriminant analysis (DA) [159] can also be used to reduce the feature dimension.”
• A survey of content-based image retrieval with high-level semantics (2007, background): “Though some algorithms [88,89] have been designed to discrete continuous attributes, whether these generally designed algorithms can always provide meaningful splitting of image feature space is a question.”
• Lazy Decision Trees (1996, methodology): “We chose to discretize the instances using recursive minimization of entropy as proposed by Fayyad & Irani (1993) and as implemented in MLC++ (Kohavi et al. 1994)) which is publicly available and thus allows replication of this discretization step.”
• Feature Generation Using General Constructor Functions (2002, methodology): “…is nominal or orderd-nominal: Interval(f1) = fIs(f1; fCig) j Ci 2 Range(f1)g where the builtinconstructor Is is de ned as Is(f1; Ci) = TRUE , f1 = Ci. f1 is continuous: Let fC1; C2 : : :Cng be a discritization of Range(f1) (We use the dynamic discretiza-tion algorithm by Fayyad and Irani (1993).).”
• A weakly informative default prior distribution for logistic and other regression models (2008, methodology): “We also use the Fayyad and Irani (1993) method for converting continuous predictors into discrete ones.”
• Discretization: An Enabling Technique (2002, background): “In Fayyad and Irani (1993), they propose that potential cutpoints are those that form boundaries between classes after sorting the continuous feature values.”
• Efficient Feature Selection via Analysis of Relevance and Redundancy (2004, methodology): “A new feature selection algorithm FCBF is implemented and evaluated through extensive experiments comparing with three representative feature selection algorithms.”
• Using Neural Network Rule Extraction and Decision Tables for Credit - Risk Evaluation (2003, methodology): “All inputs are discretized using the discretization algorithm of Fayyad and Irani (1993) with the default options on the training set.”
• Markov blanket-embedded genetic algorithm for gene selection (2007, methodology): “Because the C-correlation measure is not designed for continuous data, we use Fayyad and Irani’s Minimum Description Length (MDL) method [45] to discretize the microarray data as a pre-processing step.”
• Not So Naive Bayes: Aggregating One-Dependence Estimators (2005, methodology): “In contrast, under approaches such as naive Bayes where there is no choice in the form of the model, all that changes when the training data changes is the underlying conditional probability tables which tends to result in relatively gradual changes in the pattern of classification.”
• Building comprehensible customer churn prediction models with advanced rule induction techniques (2011, background): “All continuous variables are discretized following Fayyad and Irani (1993).”