Options

the WEIGHT TYPE?

fritmorefritmore Member Posts: 90 Contributor II
edited November 2018 in Help
Hi there

how to use/benefit from using Type WEIGHT?
There is no info about it in the RM 5 manual.
Why using Weight type at all? I mean why not just using nominal or integer?

Or is it ment for ICA model weights, NN weights...?

thx

Answers

  • Options
    landland RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 2,531 Unicorn
    Hi,

    there are several things called "weight" in RapidMiner. There is the Object "Attribute Weights" which contains weights per column (=attribute) in a dataset. And there's the special role "weight" for an attribute that means that this attribute expresses the weight or importance of the particular example (= row) of the data set.

    Greetings,
    Sebastian
  • Options
    fritmorefritmore Member Posts: 90 Contributor II
    Sebastian Land wrote:

    Hi,

    there are several things called "weight" in RapidMiner. There is the Object "Attribute Weights" which contains weights per column (=attribute) in a dataset. And there's the special role "weight" for an attribute that means that this attribute expresses the weight or importance of the particular example (= row) of the data set.

    Greetings,
    Sebastian
    Hi Sebastian

    but how is it treated by different models (role weight).
    Is it something like prior knowledge in bayesian modeling? It clearly doesnt make any sense in e.g. NeuralNet operator, kmeans........

    It is weird there is no detailed documentation for each type especially these non obvious ones like the Weight.


  • Options
    landland RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 2,531 Unicorn
    Hi,
    this hasn't anything to do with prior knowledge or something like this. The weight only expresses the importance of a certain example. If one example has a weight of 2 then the result of an analysis will be the same as if the example occurs twice with the same attribute values and a weight of 1.

    You can use this for example to aggregate your data and keep the relative importance of the aggregated examples. If 10 examples have been aggregated to one row, this row should have a weight of 10 to distinguish it's importance from any row that hasn't been aggregated.

    Greetings,
      Sebastian
  • Options
    fritmorefritmore Member Posts: 90 Contributor II
    Sebastian Land wrote:

    Hi,
    this hasn't anything to do with prior knowledge or something like this. The weight only expresses the importance of a certain example. If one example has a weight of 2 then the result of an analysis will be the same as if the example occurs twice with the same attribute values and a weight of 1.

    You can use this for example to aggregate your data and keep the relative importance of the aggregated examples. If 10 examples have been aggregated to one row, this row should have a weight of 10 to distinguish it's importance from any row that hasn't been aggregated.

    Greetings,
      Sebastian
    Hi Sebastian,
    got it! (i hope)  ;D

    so is this count aggregation for example benefiting model learning speed and lessens memory usage?

    Why arent these important things documented? ???
    thx

    p.s.: I just tried the aggregate operator, it looks it can be used to generate such sets as you gave as an example
  • Options
    landland RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 2,531 Unicorn
    Hi,
    did you actually read the manual? Isn't it mentioned in there?

    Greetings,
    Sebastian
Sign In or Register to comment.