Compumine - Forums

Message boards

Threads [ Previous | Next ]

rule definition

ANDREW LADIS
Rank:
Posts: 3
Joined: 9/12/06

rule definition | 9/17/06 12:49 PM

Dear sirs

We are trying to evaluate a collection prediction from dataset

the data set is like..
code last action code initial paid persentage
9100921 ΑΔΥΝΑΤΗ ΕΠΑΦΗ $123,00 $0,00 0%
8987873 ΑΔΥΝΑΤΗ ΕΠΑΦΗ $197,00 $0,00 0%
8527120 ΑΔΥΝΑΤΗ ΕΠΑΦΗ $103,50 $0,00 0%
8963347 ΑΔΥΝΑΤΗ ΕΠΑΦΗ $172,00 $0,00 0%
8973113 ΕΠΙΣΤΡΟΦΗ ΑΝΑΚΛΗΣΗ $194,50 $68,50 35%
9010542 ΑΡΝΗΣΗ ΠΛΗΡΩΜΗΣ $896,00 $0,00 0%
8884876 ΕΠΙΣΤΡΟΦΗ ΑΝΑΚΛΗΣΗ $173,50 $173,50 100%
8715715 ΕΠΙΣΤΡΟΦΗ ΑΝΑΚΛΗΣΗ $84,50 $84,50 100%
9109994 ΕΠΙΣΤΡΟΦΗ ΑΝΑΚΛΗΣΗ $149,00 $107,50 72%

The code is debtors code id the last action was what hapened when i tried to reach him in phone the initial was what be in debt the paid is what the debtor has paid and persentage is the persentage of paid of initial debt.

The goal is to predict the persentage of payments of a furture debtors data set , according to intervals of initial money.What is the prdeict persentage of payment between 70 and 100 $ pe??and so on etc etc

Is that possible or not.

Another goal is to predict what the action code will be

Keep in mind that we have a lot of data for training about 8000 rows at least

Would you please help us a litle in wich model is more preferable and things like those?

Thank you in advance

Andrew Ladis

Advisor@ath.forthnet.gr

Compumine Support
Rank:
Posts: 19
Joined: 9/8/06

RE: rule definition | 9/18/06 7:08 PM as a reply to ANDREW LADIS.

Dear Andrew,

As I understand, the target variable is "persentage". What you need to do in order to prepare the data is:

* Remove the dollar signs from the values in 'initial' och 'paid' columns.
* Remove the '%' sign from the 'persentage'-column.

Then:

* Save the data as a UTF-encoded text file.
* Import it into RDS as modeling data.
* Open the data grid for the data set by clicking the hyerlink 'Modeling data set'
* Change data type of first column ('code') to 'id'.
* Change data type of last column ('percentage') to 'regression' - if you would like to be able to obtain a score. I could also be possible to divide the values of this column into discrete intervals. Do this by adding a new column before importing the data inte RDS.
* Check that columns 'last action' and 'code' (the second one, which should be imported as 'code_0') are of type 'categoric' (if you would think that they are of any value, or you want to use them) and that the 'initial' column is of the type 'numeric'.

If you are primarily interested in making good predictions, we would suggest you use an ensemble model. The default settings are a good starting point. If you are also interested in rules, you could add a metod for creating trees, and the rule set method.

Good luck!

/Compumine Support

ANDREW LADIS Rank: Posts: 3 Joined: 9/12/06	RE: rule definition \| 9/19/06 10:29 AM as a reply to Compumine Support. Dear sirs I want to thank you very much for your help. really apreciate it! I will try your suggestion and i will inform you asap. My Best regards Andrew Ladis

Compumine Support Rank: Posts: 19 Joined: 9/8/06	RE: rule definition \| 9/19/06 7:44 PM as a reply to ANDREW LADIS. Great! Please let us know how things are going. /Compumine Support