computing brief..

madnormski · Contributor Joined: 06 Dec 2006 Posts: 6

build and compare 2 different predictive models using a decision tree..with the winconsin breast cancer dataset...

i am seriously lost on this?...

editor · Posted: Fri Dec 08, 2006 2:19 pm Post subject: Wisconsin Breast Cancer Data

You can get this data from
UCI repository - see
http://www.ics.uci.edu/~mlearn/MLSummary.html

Look at
http://www.ailab.si/orange/doc/datasets/breast-cancer-wisconsin.htm
for notes and analysis

madnormski · Contributor Joined: 06 Dec 2006 Posts: 6

i have the dataset...how do i create 2 predictive models using decision trees using clementine..im more into multimedia projects and i am lost on this...?????????

TimManns · Posted: Sun Dec 10, 2006 6:04 pm Post subject:

Hi,

When installing Clementine you have the choice to include small demo files and demo Clementine stream files that provide examples of a simple medical drup prescription scenario. This may be similar to your breast cancer project. Following a simple example from the appendix of the User guide (installed as PDF when you install Clementine) will show how to build simple predictive models. It will cover the basics such as balancing the data, sampling, and simple data trasnformations prior to building models.

Clementine is easy to use, but you should expect some sort of hands-on training from a teacher or colleague though - at least for an hour or so.

Cheers

Tim

TimManns · Posted: Sun Dec 10, 2006 6:07 pm Post subject:

I should have mentioned;
- it could take you only 30 mins if you know the data and are familar with Clementine to create a basic predictive model.

madnormski · Contributor Joined: 06 Dec 2006 Posts: 6

do u happen to know wot theoutput attribute is for the wisconsin dataset?..

editor

[quote="madnormski"]do u happen to know wot theoutput attribute is for the wisconsin dataset?..[/quote]

It is the class - the last attribute, (benign or malignant)

madnormski · Contributor Joined: 06 Dec 2006 Posts: 6

im 30 yaers old from derry in northern ireland...if i sent you the breast cancer dataset could you design 2 predictive models for me?...plzzzzz...im lost completely,,,

TimManns · Posted: Wed Dec 13, 2006 5:21 pm Post subject:

sure- i can spare half an hour to hack something togther.

It would help if you could supply information about the data, the business case or any background information. I'd definately need a description of the variables/fields involved. Simply having field names isn't going to be enough, I'll need to know what the data actually is.

Cheers

Tim
timmanns at bigpond dot net dot au

madnormski · Contributor Joined: 06 Dec 2006 Posts: 6

tim could i possibly email you as i could send u the small dataset..

i have to build and compare 2 different predictive models using a decision tree, a neural network or cohonen network node in clementine...

i have to pre process and clean the data and then properly define input and output variables to achieve a satisfactory solution to problem...

finaallly..

ihave to write up a report..
-including problem domain
-input, output and excluded variables (and reasons to designate them as such)
- explain choice of algorithms and of data pre processing activity..
-give an analysis of results including an assessment of effectiveness of the mining...

my email is ...normanshongo2003@yahoo.co.uk

tom i will never be using this program again as i am a multimedia student and have no idea on wot to do for this small assignment..

can i send u the dataset files plz?...

could u include the aspects of the project brief...maybe screenshots from clementine with small notes included....anything at all would be greatly appreciated as i have to hand this in tomorrow evening....as long as the topics in brief covered it does not have to be a very indept investigation....

tom u would be really helping me out as i dont know who to turn too now and i want to get the brief completed in time...

thank you so much...if u could do a half houts piece i would be so happy...[b][/b]