!!This is "autoread.txt"  file. I you do not wish to load this file automatically, just remove or rename 
it in your directory!!

REGRT is the program for analysis of mutational spectra using decision trees 
(for details, see: 
Berikov V.B., Rogozin I.B.  "Regression trees for  analysis of mutational spectra in nucleotide sequences".
Bioinformatics, vol. 15, 1999. pp.553-562).

To analyse your data, you should:

1. Open your data file (File|Open). 
This file should be written in the following format, for example  (the text between symbols # is a 
comment ignored by program):

   22    # number of objects (cases); in the given version - no more than 1000#
   16    #number of variables; in the given version - no more than 50# 
    N   #code of unknown value#

 #types of features: 1 - numerical; 0(M) - categorical,  M =number of categories (classes)#
  1  0(4)  0(4)  0(4)  0(4)  0(4)  0(4)  0(4)  0(4)  0(4)  0(4)  0(4)  1  1  1  1

 #feature names:#
Y X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15

#data table: (use space or ',' between different values; decimal values are denoted by '.'#
  0  1  2  3  2  4  3  4  1  3  1  3  3  2  4  2
  2  2  2  4  4  4  3  4  3  2  3  3  0  3  4  4
  0  4  4  4  3  4  3  2  3  3  2  3  0  2  5  4
  4  1  1  1  1  1  3  2  3  3  1  1  7  1  3  0
 14 3  4  3  2  3  3  4  1  4  1  1  3  1  4  3
15  3  3  4  3  3  3  4  1  1  1  4  3  0  5  3



You can edit this file using built-in editor.

2. Set parameters of algorithm (Run|Parameters)

Recursive complexity: 
This parameter defines maximal depth of search of optimal
regression tree.  The integer part of rev defines guaranteed
size (r) of the search depth. The fraction part defines a portion
of the most informative variables (i.e. which give the best prediction), 
which are taken for the search of r+1,r+2,... depth
(from variables, taken on previous stage). Note: large values of this parameters makes the program 
time- anf memory consuming;

Homogeneity: 
Critical value for X2 statistics. The node does not need further branching if the objects in 
the node are homogenious.

VariantsNumber: 
Desired number of regression tree variants. The variants are built by the way of consequent excluding
the most informative (corresponding to the root node) variable in the preceding variant. 
If test sample is given, the variants are used for final decision design by voting procedure taking
into account the competence of each tree variant. The competence estimations are inversly
proportional to crossvalidate error for each tree.

FeatureToPredict: the number of prediction feature in data table;

FoldsCrossValidate: 
The number (L) of  folds in crossvalidation procedure.  In this procedure, 
data table is split on L random subsets, and then turn by turn each subset is used as control sample 
(the prediction error is calculated), while another parts as learning sample. 
Finally, the prediction errors are averaged over these L trials. The mean error is more 
adequate  (more close to true prediction error). Setting L=0 makes program working without crossvalidation.
In the case of regression tree, the error (standard deviation) is normalized to the standard deviation 
of the initial sample.

3. Build decision tree (Run|Grow Tree)

4. The results are automatically saved in 'output.txt' file. 

