source: prdatasets/musk1.m @ 150

%MUSK1 Prediction of molecules to be musks or non-musks
%PRTools dataset import from UCI, 476 objects, 166 features, 2 classes
%
%  A = MUSK1
%
%DESCRIPTION
%This command downloads one of the UCI data sets, converts it into PRTools
%format and stores it locally for future use. Consult the <a href="http://archive.ics.uci.edu/ml/datasets/Musk+(Version+1)">related website</a>
%for further information. Please make the appropriate references in
%publications that make use of this dataset.
%
%A is a multiple-instance dataset. The objects are the instances. They
%group in 92 bags. There are two labels: (1) the Bag Label and (2) the Bag
%Index. Use the CHANGELABLIST command to change from labeling system. See
%also MULTI_LABELING
%
%SEE ALSO <a href="http://prtools.tudelft.nl/prtools/">PRTools Guide</a>, <a href="http://archive.ics.uci.edu/ml/">UCI Website</a>
%DATASETS, MULTI_LABELING, CHANGELABLIST

function a = musk1

% UCI data is Z-compressed. No Matlab command foud for uncompressing.
% Load zipped data from prtools website.
a = pr_download_uci('Musk+(Version+1)','http://prtools.tudelft.nl/prdatasets/musk1.zip');
a = setlablistnames(a,'Bag Label',1);
a = feat2lab(a,1,'Bag Index');
a(:,2) = [];
a = setname(a,'Musk version 1');