A Bayes Evaluation Criterion for Decision Trees

Voisine, Nicolas; Boullé, Marc; Hue, Carine

doi:10.1007/978-3-642-00580-0_2

Cited by 7 publications

(3 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A decision tree is a graph which can be used as a model of a categorical variable. A decision tree aims at predicting a categorical (numerical or linguistic) output variable from a set of numerical or linguistic input variables [11]. Decision trees are useful in solving classification and prediction problems [12,13].…”

Section: Decision Treementioning

confidence: 99%

Graph-Based Decision Making in Industry

Kutschenreiter-Praszkiewicz¹

2018

Graph Theory - Advanced Algorithms and Applications

View full text Add to dashboard Cite

Decision-making in industry can be focused on different types of problems. Classification and prediction of decision problems can be solved with the use of a decision tree, which is a graph-based method of machine learning. In the presented approach, attribute-value system and quality function deployment (QFD) were used for decision problem analysis and training dataset preparation. A decision tree was applied for generating decision rules.

show abstract

Section: Decision Treementioning

confidence: 99%

Graph-Based Decision Making in Industry

Kutschenreiter-Praszkiewicz¹

2018

Graph Theory - Advanced Algorithms and Applications

View full text Add to dashboard Cite

show abstract

“…The optimal model M is then the one with the least cost c (see original work [5] for explicit expression of p(M ) and p(D|M ) and for the optimization algorithm). The generic modl approach has also already been successfully applied to supervised value grouping [4] and decision tree construction [28]. In each instantiation, the modl method promotes a trade-off between (1) the fineness of the predictive information provided by the model and (2) the robustness in order to obtain a good generalization of the model.…”

Section: Towards Modl Rulesmentioning

confidence: 99%

A Bayesian Approach for Classification Rule Mining in Quantitative Databases

Boullé

2012

Machine Learning and Knowledge Discovery in Databases

Self Cite

View full text Add to dashboard Cite

Abstract. We suggest a new framework for classification rule mining in quantitative data sets founded on Bayes theory -without univariate preprocessing of attributes. We introduce a space of rule models and a prior distribution defined on this model space. As a result, we obtain the definition of a parameter-free criterion for classification rules. We show that the new criterion identifies interesting classification rules while being highly resilient to spurious patterns. We develop a new parameter-free algorithm to mine locally optimal classification rules efficiently. The mined rules are directly used as new features in a classification process based on a selective naive Bayes classifier. The resulting classifier demonstrates higher inductive performance than state-of-the-art rule-based classifiers.

show abstract

“…Simple models such as Bernoulli or mainly multinomial distributions are important because they are easier to analyze theoretically and useful in many applications. For example, the multinomial distribution has been used as a building block in more complex models, such as naive Bayes classifiers (Mononen and Myllymäki, 2007), Bayesian networks (Roos et al, 2008), decision trees (Voisine et al, 2009) or coclustering models (Boullé, 2011;Guigourès et al, 2015). These models involve up to thousands of multinomials blocks, some of them with potentially very large numbers of occurrences and outcomes.…”

Section: Introductionmentioning

confidence: 99%

Revisiting enumerative two-part crude MDL for Bernoulli and multinomial distributions (Extended version)

Boullé¹,

Clérot²,

Hue³

2016

Preprint

Self Cite

View full text Add to dashboard Cite

We leverage the Minimum Description Length (MDL) principle as a model selection technique for Bernoulli distributions and compare several types of MDL codes. We first present a simplistic crude two-part MDL code and a Normalized Maximum Likelihood (NML) code. We then focus on the enumerative two-part crude MDL code, suggest a Bayesian interpretation for finite size data samples, and exhibit a strong connection with the NML approach. We obtain surprising impacts on the estimation of the model complexity together with superior compression performance. This is then generalized to the case of the multinomial distributions. Both the theoretical analysis and the experimental comparisons suggest that one might use the enumerative code rather than NML in practice, for Bernoulli and multinomial distributions.

show abstract

A Bayes Evaluation Criterion for Decision Trees

Cited by 7 publications

References 10 publications

Graph-Based Decision Making in Industry

Graph-Based Decision Making in Industry

A Bayesian Approach for Classification Rule Mining in Quantitative Databases

Revisiting enumerative two-part crude MDL for Bernoulli and multinomial distributions (Extended version)

Contact Info

Product

Resources

About