A Comparative Study on Feature Selection in Text Categorization
Citations
[...]
13,246Â citations
8,658Â citations
7,539Â citations
3,517Â citations
Cites methods from "A Comparative Study on Feature Sele..."
...Earlier comparisons for text classification using ranking methods can be found in [22]....
[...]
...Since we are approximating the PDF of a single feature and the output class distribution, calculation of MI will not be accurate and is easily influenced by marginal densities [22]....
[...]
3,123Â citations
References
[...]
13,246Â citations
"A Comparative Study on Feature Sele..." refers methods in this paper
...2 Information gain (IG) Information gain is frequently employed as a termgoodness criterion in the eld of machine learning[17, 14]....
[...]
4,272Â citations
"A Comparative Study on Feature Sele..." refers background in this paper
...3 Mutual information (MI) Mutual information is a criterion commonly used in statistical language modelling of word associations and related applications [7, 2, 21]....
[...]
3,571Â citations
"A Comparative Study on Feature Sele..." refers background or methods in this paper
...We also used the SMART system [18] for uni ed preprocessing followed feature selection, which includes word stemming and weighting....
[...]
...100%, the system assigns in decreasing score order as many categories as needed until a given recall is achieved, and computes the precision value at that point[18]....
[...]
...2 Experimental settings Before applying feature selection to documents, we removed the words in a standard stop word list[18]....
[...]
2,672Â citations
"A Comparative Study on Feature Sele..." refers background in this paper
...Hence, the 2 statistic is known not to be reliable for low-frequency terms[6]....
[...]
1,713Â citations
"A Comparative Study on Feature Sele..." refers background in this paper
...A recent theoretical comparison, for example, was based on the performance of decision tree algorithms in solving problems with 6 to 180 features in the native space[10]....
[...]