The CN2 Induction Algorithm

By Peter Clark, Tim Niblett. 1989

How researchers see this paper

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

In Statistical Learning Theory

Citations in the corpus

Interpretable Machine Learning: Fundamental Principles and 10 Grand Challenges (2021, methodology): “Associative classification methods assemble decision lists or decision sets from a set of pre-mined rules, generally either by greedily adding rules to the model one by one, or simply including all “top-scoring” rules into a decision set, where each rule is scored separately according to a scoring f”
Interpretable Two-level Boolean Rule Learning for Classification (2015, methodology): “The second formulation replaces the 0-1 classiﬁcation error cost with the Hamming distance from the current rule to the closest rule that correctly classiﬁes a sample; we propose block coordinate descent and alternating minimization approaches for optimizing the objective in the second formulation.”
Learning Certifiably Optimal Rule Lists (2017, background): “, early algorithms [9, 15, 29]) and associative classi cation [41].”
Optimal Decision Lists using SAT (2020, methodology): “Decision lists were introduced by Rivest (1987) and heuristic methods for decision lists also date back to the late 80s (Clark and Niblett 1989; Clark and Boswell 1991).”
Computing Optimal Decision Sets with SAT (2020, methodology): “A number of state-of-the-art algorithms were additionally considered including the heuristic methods CN2 [11,12], and RIPPER [13], as well as MaxSAT-based IMLI [20] (which is a direct successor of MLIC [39]).”
IMLI: An Incremental Framework for MaxSAT-Based Learning of Interpretable Classification Rules (2019, methodology): “Speciﬁcally, decision rule approaches such as C4.5 rules (Quinlan 1993), CN2(Clark and Niblett 1989), RIPPER (Cohen 1995), SLIP-PER (Cohen and Singer 1999) rely on heuristics, branch pruning, ad-hoc local criteria e.g., maximizing information gain, coverage, etc. because these models consider an int”
MLIC: A MaxSAT-Based Framework for Learning Interpretable Classification Rules (2018, background): “For example vaious popular decision rule approaches, such as C4.5.rules [29], CN2 [9], RIPPER [10], SLIPPER [11], all make different trade-offs in how they use these heuristic criteria for growing and pruning the rules.”
Boolean Decision Rules via Column Generation (2018, background): “Predominant practical approaches include a covering or separate-and-conquer strategy ([15, 14, 16, 26, 28, 40],”
Interpretable Decision Sets: A Joint Framework for Description and Prediction (2016, neutral): “Finally, we note that prior work [9] extended the CN2 rule-learning algorithm [10] to produce unordered rule sets.”
Scalable Bayesian Rule Lists (2016, methodology): “Rule learning algorithms have been developed possibly since the AQ algorithm in the 1960s (Michalski, 1969) and throughout the 1980’s and 90’s (Cendrowska, 1987; Quinlan, 1983; Clark & Niblett, 1989; Cohen, 1995; Liu, Hsu, & Ma, 1998).”
PREDICTIVE LEARNING VIA RULE ENSEMBLES (2008, methodology): “Actual ILP algorithms such as CN2 [Clark and Niblett (1989)], RIPPER [Cohen (1995)], and PRO-GOL [Muggleton (1995)] diﬀer with respect to the detailed techniques that implement the generic paradigm.”
Using Decision Lists to Construct Interpretable and Parsimonious Treatment Regimes (2015, methodology): “…of decision list, which was developed in the computer science literature for representing flexible but interpretable classifiers (Rivest, 1987; Clark and Niblett, 1989; Marchand and Sokolova, 2005; Letham et al., 2012; Wang and Rudin, 2014); see Freitas (2014) for a recent position article on…”
Finding a short and accurate decision rule in disjunctive normal form by exhaustive search (2010, methodology): “The method used by CN2 to induce an unordered rule set can possibly be applied (Clark and Boswell 1991).”
Rule Extraction from Support Vector Machines: An Overview of Issues and Application in Credit Scoring (2008, background): “Many algorithms are capable of learning rules or trees directly from a set of training examples, e.g., CN2 [15], AQ [45], RIPPER [16], AntMiner+ [44], C4.5 [49] or CART [11].”
An Extension on "Statistical Comparisons of Classifiers over Multiple Data Sets" for all Pairwise Comparisons (2008, unknown): “…sets of hypotheses considering 4 classifiers k m = k 2 2 m n e 4 6 64 14 5 10 1024 51 6 15 32768 202 7 21 2097152 876 8 28 2.7 · 10 8 4139 9 36 6.7 · 10 10 21146 Table 1: All pairwise comparisons of k classifiers NaiveBayes, Kernel (McLachlan, 2004) 1 and, finally, CN2 (Clark and Niblett, 1989).”
Data Classification: Algorithms and Applications (2014, methodology): “121 5.2 Rule Induction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 5.2.1 Two Algorithms for Rule Induction . . . . . . . . . . . . . . . . . . . . 123 5.2.1.1 CN2 Induction Algorithm (Ordered Rules) . . . . . . . . . . 124 5.2.1.2 RIPPER Algorithm and Its Variations (Or”