Profiles
Data Mining is an iterative process and it is important to be able to retrieve all the context of a session.
Profiles allow storing information concerning the first three steps of the assistant, i.e., those preceding rules computation (Step 4).
SAVING A PROFILE:
Profiles are flexible: they can store information on all the steps, or on only a given step.
Therefore, it is possible to store independently information concerning the three steps:
- the choice of the attributes to load from the database ,
- the way attributes and their modalities appear in the rule schemes,
- the choice of the datamining technique and the configuration of its parameters. Let us note that each configuration of techniques can be stored independently from the others.
Depending on the step in the assistant, the command for saving a profile opens a panel, which offers the choice of the parameters to save through a set of boxes to check.
All combinations are possible, even incomplete ones. When loading, QuantMiner completes the parameters according to the context.
It is important to give the profile files (*.prf) a name expressing their uses (especially the database to which they apply).
LOADING A PROFILE:
Whatever the first three steps of the assistant the user is concerned, he/she can load a profile storing parameters for this step, or for another (previous or following) one.
It is possible to combine profiles, for instance using a first profile giving the choice of the attributes to load from the database and another one concerning the parameters of the genetic algorithm.
REMARK: It is possible to choose the parameters to save in a profile, but it is not possible to choose the parameters to load: all the parameters present in a file are used.
If the contents of the database has changed or if some parameters are not adapted, QuantMiner does its best to load the profile, taking into account the current situation.