Verification Experiment on Feature Word Selection Method: "the influence of various Feature Word Selection Algorithms on text classification performance": The paper has a brief understanding of this issue (download of experimental data)

(Note: the author and author of the blog post: finallyliuyu source blog)

The purpose of publishing the experiment data is to facilitate users with the same fans to download intermediate data and quickly reproduce the experiment.

Resource Space Provider: download.csdn.net

Statistical dictionary and associated table data structure

Each packet contains four files: keywords.dat,testvsm.dat,trainingvsm.dat,evaluation.txt (The. dat file needs to be viewed using ultraedit)

Select 2000 keywords using the Global DF Method

Use the IG method to select 2000 keywords

Card method selects 2000 keywords

Select 4000 keywords using the local DF Method

Point mutual information method selects 2000 keywords

 

 Use the local DF method to select 1000 Feature Words

Use the global DF method to select 1000 Feature Words

Use the dot mutual information method to select 1000 Feature Words

Use the IG method to select 1000 Feature Words

Card method selects 1000 Feature Words