Multi-granularity feature selection on cost-sensitive data with measurement errors and variable costs

Liao, Shujiao; Zhu, Qingxin; Qian, Yuhua; Lin, Guoping

doi:10.1016/j.knosys.2018.05.020

Cited by 24 publications

(25 citation statements)

References 53 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In most existing cost‐sensitive attribute‐scale selection approaches, for a numeric attribute, its scale is measured by the error confidence level of the attribute values. Especially, a confidence level vector was introduced in [23] so that different numeric attributes can correspond to different confidence levels. On this basis, the confidence level vector‐based decision system and neighborhood were defined.…”

Section: Generalized Confidence Level Vector‐based Neighborhood Rough...mentioning

confidence: 99%

“…The reason that

2 e (a, p_{a})

but not

e (a, p_{a})

is used in Equation (3) has been explained in [23], and the standard deviation

σ_{a}

in Equation (4) is set to be

σ_{a} = k \cdot m a x | a (x_{i}) - \overset{a (x)}{true} | (i = 1, 2, …, | U |),

where

k > 0

is an adjusting factor,

a (x_{i})

is the attribute value of object

x_{i}

w.r.t. attribute

a

, and

\overset{a (x)}{true} = \frac{1}{| U |} \sum_{i = 1}^{| U |} a (x_{i})

is the average attribute value of

a

for all objects.…”

Section: Generalized Confidence Level Vector‐based Neighborhood Rough...mentioning

confidence: 99%

“…It can select a nice pair of attribute subset and generalized confidence level vector, namely the scale combination so that the consumed average total cost is minimized under any rational value of test cost upper bound, and it is efficient due to the use of four accelerating techniques. The results of the attribute‐scale selection algorithm can be further enhanced through using a competition strategy mentioned in [23]. Finally, detailed experiments conducted on eleven hybrid UCI data sets 27 thoroughly verify the effectiveness of the proposed cost‐sensitive attribute‐scale selection approach.…”

Section: Introductionmentioning

confidence: 98%

“…Recently several novel cost‐sensitive attribute‐scale selection approaches have been proposed based on realistic measurement errors, test costs and misclassification costs 21–24 . The motivation is that, in practical applications, due to the limitation of measuring tools or means, measurement errors often occur when acquiring attribute values; and different error ranges of attribute values result in different levels of scale.…”

Section: Introductionmentioning

confidence: 99%

“…Since the attribute‐scale selection approaches can simultaneously select attributes and scale combination, they are more comprehensive than the approaches for pure attribute reduction and pure scale selection. However, most existing cost‐sensitive attribute‐scale selection approaches assume that all attributes have the same scale, so they are not applicable in some real applications; only the approach in [23] considers the scale diversity between different attributes, so it is more versatile and practical than others. Nevertheless, the approach is only suitable for numeric data but not hybrid data, and it does not take into account the test cost constraint case where the total test cost consumed by a user cannot exceed a given upper bound.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Attribute‐scale selection for hybrid data with test cost constraint: The approach and uncertainty measures

Liao

Lin

et al. 2021

Int J of Intelligent Sys

Self Cite

View full text Add to dashboard Cite

Recently several novel cost-sensitive attribute-scale selection approaches have been proposed based on measurement errors. They are significant because they can simultaneously select attributes and scale combination to minimize the cost consumed in data processing. However, these approaches cannot deal with hybrid data with test cost constraint, and most of them do not consider the scale diversity between different attributes; and these approaches do not touch the uncertainty measurement, all of which are important issues in real applications. To address this situation, in this paper an effective cost-sensitive attribute-scale selection approach is presented based on the rough set theory, and multiple relevant uncertainty measures are developed. The main contributions of the paper are threefold. First, a generalized confidence level vectorbased neighborhood rough set model is constructed. It takes into account the scale diversity between different attributes of hybrid data. Then, multiple uncertainty measures are developed. They consider both attributes and scales, thus are more general than existing ones which consider only attributes or only scales. Finally, an efficient heuristic attribute-scale selection algorithm is designed, which can select attributes and their respective scales to minimize the consumed total cost of

show abstract

Section: Generalized Confidence Level Vector‐based Neighborhood Rough...mentioning

confidence: 99%

“…The reason that