A modified Chi2 algorithm for discretization

Tay, Francis Eng Hock; Shen, Lixiang

doi:10.1109/tkde.2002.1000349

Cited by 132 publications

(72 citation statements)

References 7 publications

(6 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Since the evaluation function in our algorithm is proposed based on the heterogeneity compared to the entropy criterion, our method would be compared with Ent-MDLC, which is a well-known representative of entropy-based discretization methods. Its good performance has often been shown in previous research [7], [12], [15], [22]. It is recommended as the first choice when other things are equal, such as user need, class information, and other considerations [15].…”

Section: Experiments Resultsmentioning

confidence: 91%

“…Our method has been compared, in terms of predictive error rate and tree size, with Ent-MDLC, the entropy-based method that is known for its good performance [7], [12], [15], [22].…”

Section: Resultsmentioning

confidence: 99%

“…Statistics-based algorithms include ChiMerge [10], Chi2 [16], Modified Chi2 [22], StatDisc [19], Khiops [2], and so forth. The ChiMerge method [10], [16], [22] searches for the best merge of adjacent intervals by minimizing the Chisquare criterion applied locally to two adjacent intervals that are merged according to statistical similarity.…”

mentioning

confidence: 99%

“…The ChiMerge method [10], [16], [22] searches for the best merge of adjacent intervals by minimizing the Chisquare criterion applied locally to two adjacent intervals that are merged according to statistical similarity. Like ChiMerge, StatDisc [19] considers merging up to N adjacent intervals at a time where N is a user-specified parameter.…”

mentioning

confidence: 99%

See 3 more Smart Citations

A discretization algorithm based on a heterogeneity criterion

Liu

Wang

2005

IEEE Trans. Knowl. Data Eng.

122

View full text Add to dashboard Cite

Abstract-Discretization, as a preprocessing step for data mining, is a process of converting the continuous attributes of a data set into discrete ones so that they can be treated as the nominal features by machine learning algorithms. Those various discretization methods, that use entropy-based criteria, form a large class of algorithm. However, as a measure of class homogeneity, entropy cannot always accurately reflect the degree of class homogeneity of an interval. Therefore, in this paper, we propose a new measure of class heterogeneity of intervals from the viewpoint of class probability itself. Based on the definition of heterogeneity, we present a new criterion to evaluate a discretization scheme and analyze its property theoretically. Also, a heuristic method is proposed to find the approximate optimal discretization scheme. Finally, our method is compared, in terms of predictive error rate and tree size, with Ent-MDLC, a representative entropy-based discretization method well-known for its good performance. Our method is shown to produce better results than those of Ent-MDLC, although the improvement is not significant. It can be a good alternative to entropy-based discretization methods.

show abstract

Section: Experiments Resultsmentioning

confidence: 91%

“…Our method has been compared, in terms of predictive error rate and tree size, with Ent-MDLC, the entropy-based method that is known for its good performance [7], [12], [15], [22].…”

Section: Resultsmentioning

confidence: 99%

mentioning

confidence: 99%

mentioning

confidence: 99%

See 2 more Smart Citations

A discretization algorithm based on a heterogeneity criterion

Liu

Wang

2005

IEEE Trans. Knowl. Data Eng.

122

View full text Add to dashboard Cite

show abstract

“…In the bottom-up category, there are widely used algorithms such as ChiMerge [21], Chi2 [22], Modified Chi2 [23], and Extended Chi2 [24]. Bottom-up method starts with the complete list of all continuous value of the attribute as cut-points, so its computational complexity is usually higher than the top-down method [29].…”

Section: Related Workmentioning

confidence: 99%

A Discretization Algorithm for Uncertain Data

Xia

2010

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. This paper proposes a new discretization algorithm for uncertain data. Uncertainty is widely spread in real-world data. Numerous factors lead to data uncertainty including data acquisition device error, approximate measurement, sampling fault, transmission latency, data integration error and so on. In many cases, estimating and modeling the uncertainty for underlying data is available and many classical data mining algorithms have been redesigned or extended to process uncertain data. It is extremely important to consider data uncertainty in the discretization methods as well. In this paper, we propose a new discretization algorithm called UCAIM (Uncertain Class-Attribute Interdependency Maximization). Uncertainty can be modeled as either a formula based or sample based probability distribution function (pdf). We use probability cardinality to build the quanta matrix of these uncertain attributes, which is then used to evaluate class-attribute interdependency by adopting the redesigned ucaim criterion. The algorithm selects the optimal discretization scheme with the highest ucaim value. Experiments show that the usage of uncertain information helps UCAIM perform well on uncertain data. It significantly outperforms the traditional CAIM algorithm, especially when the uncertainty is high.

show abstract

Discretization

2015

Data Mining Algorithms

View full text Add to dashboard Cite

A modified Chi2 algorithm for discretization

Cited by 132 publications

References 7 publications

A discretization algorithm based on a heterogeneity criterion

A discretization algorithm based on a heterogeneity criterion

A Discretization Algorithm for Uncertain Data

Discretization

Contact Info

Product

Resources

About