
This package include two folder: "TestInfo" and "RFModel".
          
	"TestInfo": Experiment for Figure 1 in Section 5.3 of our paper.
        "RFModel": Experiment for Table 3 and Figure 2,3,4,5 in Section 5.3 and 5.4 of our paper.




1) ***Datasets***: 

Nine datasets: Fbis, Gisette, La1s, La2s, Mnist, Newsgroups, Re1, Tis, Wap


  


2)***File explaination***:

	(1) data file (C4.5 files formats): *.data, *.test. *.names (e.g. mnist.data, mnist.test, mnist.names)
	(2) executable progrom: 
             "TestInfo.exe" is executable program for Figure 1 in Section 5.3 of our paper.
             "RandomForest.exe" is executable program for Table 3 and Figure 2,3,4,5 in Section 5.3 and 5.4 of our paper.
	(3) Batch file: *.bat (e.g. Mnist_RF_CS.bat, run_Mnist_Info.bat)

            Explanation for command Line in batch files:
            for examle, "run_Mnist_Info.bat", Mnist_RF_"Mnist_RF_Old.bat", "Mnist_table.bat", "Mnist_RF_CS.bat", "Mnist_RF_IGR.bat", 

            "run_Mnist_Info.bat"----"TestInfo -f mnist -s 1 -i 0.0015" 
            ("-s 1 " stand for Chi-square method; "-s 2 " stand for Information gain ratio method;)
            ("-i 0.0015" stand for informativeness threshold)

            "Mnist_RF_Old.bat"------"RandomForest -f mnist -k 100 -a 1 -d 5 -r 1 -u"
            ("-k 100 " stand for the number of build tree)
            ("-a 1 " stand for Breiman origin method <"1">)
            ("-d 5 " stand for subspace size <size is equal to "5">)
            ("-r 1 " stand for random seed <seed is equal to "1">)

            "Mnist_RF_CS.bat"------"RandomForest -f mnist -k 100 -a 2 -d 5 -r 1 -u"
            ("-k 100 " stand for the number of build tree)
            ("-a 2 " stand for our weight method <chi-square statistic "2"> )
            ("-d 5 " stand for subspace size <size is equal to "5">)
            ("-r 1 " stand for random seed <seed is equal to "1">)

            "Mnist_RF_IGR.bat"------"RandomForest -f mnist -k 100 -a 3 -d 5 -r 1 -u"
            ("-k 100 " stand for the number of build tree)
            ("-a 3 " stand for our weight method <information gain ratio "3">)
            ("-d 5 " stand for subspace size <size is equal to "5">)
            ("-r 1 " stand for random seed <seed is equal to "1">)


            May be you want to ask for.....whats meaning:
            "RandomForest -f mnist -k 100 -a 3 -d 5 -r 1 -u"
            "RandomForest -f mnist -k 100 -a 3 -d 5 -r 2 -u"
               ........
            "RandomForest -f mnist -k 100 -a 3 -d 5 -r 80 -u"

            This meant we built 80 random forest model by each algorithm and calculated the average values of the four measures Strength, Correlation, error bound, Accuracy.







3)***To execute our experiment, you have to make preparations includes two things***: 

	(1)Hardware environment: 
	   Two executable program are coded in C and running on a Intel(R) Xeon(R) E5310 with 1.6GHz (8 Cores) and 8GB memory. 

	(2)Software environment: 
	   Windows Xp or Windows Server 2003 (OS); 
           Visual C++ Redistributable (vcredist_x86.exe); 
           OriginPro 8.0 (software for drawing picture); 
           UltraEdit-32 (Software for text editor);






4) ***Execute program***:

   Double-click "*.bat" file (e.g."run_Mnist_Info.bat"). At the end of executing program, OS create a file "*.out" (e.g., mnist.out)

   Explanation for the experimental results in "*.out" file:

   (1)When you execute program("TestInfo.exe"), you will see (e.g., "Proportion: 0.264103"), the digits "0.264103" is the computation result (proportion of informative features).

   (2)When you execute program("RandomForest.exe"), you will see (e.g., "5	100	0.605640	0.232836	0.634778	0.890182"), "5" is subspace size, "100" is the number of building tree, "0.605640" is Strength, "0.232836" is correlation, "0.634778" is error bound, "0.890182" is test accuracy.

   If you get the results of Table 3 in Section 5.3, you may see the results of Figure 4,5 according to the number of features is "log2(M)+1".





5)***An explanation on how to reproduce the results that you have obtained***:

To draw picture as Figure 1,2,3,4,and 5, you have to copy the result of execution to <OriginPro 8.0>. Here I am not referring to use <OriginPro 8.0> to how to drawing picture.





******If you have any questions, please contact me! Email:amusing002@gmail.com******