Projects > Discovery Challenge 2004 > Data transformation > Computing further characterstics


TimeTransf module transformations

The goal of transformations made by TimeTransf module is to calculate values (so called characteristics) for each time series in matrix Contr. A characteristics is for example an average weight of a patient. You can check a complete list of preprocessed characteristics. The final preprocessed characteristics are ready for download and use.

It is possible to create your own characteristics by using TimeTransf module. This page shows you how. As an examaple how the characteristics are defined in TimeTransf you can download the metabase where are already prepared.

Preparing data sources

TimeTransf module is a part of LISp-Miner system, which uses ODBC for database access. You can use any OBDC capable database even when we have described only MS Access as an example. Please look in the LISp-Miner documentation for further help on how to setup data sources.

Step by step procedure:

  1. Download LISp-Miner and run LM Admin.
  2. Select either empty metabase or the one prepared by us.
  3. Name the data sources as you wish and press OK
  4. Run LM Data source, click on Data matrices list and select matrix Control (this must be done to make LISp-Miner register the matrix Control)
  5. Now you can use LISp-Miner including TimeTransf module.

Using TimeTransf

If you have decided to use metabase prepared by us, everything is ready for geting results. In other cases follow further instructions:

Create variables

Variable includes information for complex description of the time series:

Type of Information Column Name in “Control” Matrix Description
Key item relation ICO identification number of patient, this data field relates directly to key items field in key table
Field for date/time Date the time information about individual events in time series, you can use ROKVYS column alternatively
Value variable various attributes the actual numeral values of events

To create a new variable click on variables icon on the task bar. For further help look into TimeTransf module documentation w_variables. A variable can be for example a weight of a patient (attribute HMOT). If you use the function average you get the average weight for each patient.

Create TimeTransf task

Characteristics are organized within tasks. Characteristics within a particular task are calculated at the same time and the final results are written to one particular result matrix. In the prepared metabase is created one task named “PKDD2003” where are defined all the characteristics described below.

To create a new task click on new task icon or change the task by double clicking on the name of the task.

Task detail window:

Creating characteristics

Once you are editing a task (see the picture above) you can create new characteristics by clicking on the Add button. The characteristics detail window appears:

Setup your new characteristics as you need. Note especially the name of characteristics (this is the name of a column in the result matrix), the function (for example Average) and variable (for example weight – HMOT). For further help look into TimeTransf documentation.

Run the transformation computing

Once all characteristics are created click on Run button. The process may take a while. Result matrix is saved along your original analyzed data. The name of the matrix is defined in Task detail window. You can view this matrix by clicking on Results button.

To use this new preprocessed data, run LM DataSource to register this new matrix. Then you are ready to use it within LISp-Miner.

Printable version of page    PDF version of page

Projects > Discovery Challenge 2004 > Data transformation > Computing further characterstics


Mail to: webmaster