Normalization of the first 10 attributes and see if it is useful or not
Discuss the relation between the number of training instances and the accuracy.
For easier analysis, you are allowed to use any faster tools.
Two recent studies on this data set are
@Article{RC02a,
author = {Collobert, R. and Bengio, S. and Bengio, Y.},
title = {A Parallel Mixture of {SVMs} for Very Large Scale Problems},
journal = {Neural Computation},
year = {2002},
volume = 14,
number = 05,
pages = {1105--1114},
ipdmembership="learning",
ipdpostscript="ftp://ftp.idiap.ch/pub/reports/2001/rr01-12.ps.gz",
IPDpdf ="ftp://ftp.idiap.ch/pub/reports/2001/rr01-12.pdf",
ipdxref = "techreport:collobert-rr-01-12.bib"
}
@Article{XL04a,
author = {Xiaomei Liu and Lawrence O. Hall and Kevin W. Bowyer},
title = {Comments on ``{A Parallel Mixture of SVMs for Very Large Scale Problems}''},
journal = {Neural Computation},
year = {2004},
OPTkey = {},
volume = {16},
number = {7},
pages = {1345--1351},
OPTmonth = {},
OPTnote = {},
OPTannote = {}
}