pnookala / cs584_project Goto Github PK

/* Authors: Karl Stough, Poornima Nookala */

This document contains instructions about how to run the program.

Requirements: 
	Python 3.5
	Python Libraries:
		matplotlib
		numpy
		scipy
		sklearn

There is a help menu which displays information about what the program can accept. Here are the details of command line options available to run the program.

project_main.py -i <input_file> [-l <class label column>] [-s] [-r <row_rand>] [-c <col_rand>] [--seed] [--skip-columns] [--no-sort]
    -l  -    The column containing the class label. Defaults to the last column [0-<# columns>]
    -s  -    skips the first line of files that have a header
    -r  -    Parameter for clearing data, the rate at which rows should be zeroed [0-1]
    -c  -    Parameter for clearing data, the number of columns to potentially clear [1-<#features>]
    --seed - The random seed to use in clearing the data [Numeric]
    --skip-columns - Specify which columns should be ignored during the classification step (0-indexed)
    --no-sort - Tells the imputation algorithm to impute on the unordered set of missing values

Both the -r and -c parameters will accept a triplet of values specifying a range (as [start,end,step])
for the given parameter. The resultant range must still lie within the arguments' bounds
The --skip-columns parameter accepts a comma-delimited set of columns

Program can handle numerical as well as categorical features. Any new datasets should be copied to 'data' folder to run the program.
The 'results' folder contains results from experiments performed by the team.

==============================================================================================================================================================

Steps to run the program
-------------------------

1. Navigate to project/code folder from terminal.
2. runtests.sh contains all the commands to run the program for 7 different datasets which are included in the data folder.
3. From terminal, execute 'sh runtests.sh'.
4. All the results and plots of the program will be saved in 'output' folder.

==============================================================================================================================================================