regression methods clustering methods then LIME Method
asiddiq
New Altair Community Member
I need to get an example of how to Design regression methods with some clustering methods than using Local Interpretation (LIME)?
Tagged:
1
Best Answers
-
Hi @asiddiq , you could use the RapidMiner operators to manipulate the data so that they are in the way you want and then use the "Execute R" operator to execute LIMEWhat clustering methods do you want to apply?
5 -
Hi think he is referring to the operator i wrote a while ago called Local Interpretation (LIME). This does not support clustering. Maybe you want to use Explain Predictions, which is very similar.Best,Martin
5 -
@asiddiq
Are you trying to use regression clustering together for predictions ? This can achieved by utilizing Hierarchical classification, wherein a clustering algorithm breaks your data into clusters and you have regression models for each clusters. Thus, ensuring different sets of data have different models.
For model interpretability, you can use Explain Predictions as @mschmitz suggested, the operator can suggest global variable importance and impact on model base output, as well as, local results of attributes supporting and contradicting the model output for every row. You can also check out SHAP in python, which is a game theory based model explanation package.
1
Altair Employee
Answers
-
Hi @asiddiq , you could use the RapidMiner operators to manipulate the data so that they are in the way you want and then use the "Execute R" operator to execute LIMEWhat clustering methods do you want to apply?
5 -
Hi think he is referring to the operator i wrote a while ago called Local Interpretation (LIME). This does not support clustering. Maybe you want to use Explain Predictions, which is very similar.Best,Martin
5 -
I will try multiple algorithms to test the best one in my prediction model as well as using Local Interpretation (LIME) in the end to explain the outputs0
-
@asiddiq
Are you trying to use regression clustering together for predictions ? This can achieved by utilizing Hierarchical classification, wherein a clustering algorithm breaks your data into clusters and you have regression models for each clusters. Thus, ensuring different sets of data have different models.
For model interpretability, you can use Explain Predictions as @mschmitz suggested, the operator can suggest global variable importance and impact on model base output, as well as, local results of attributes supporting and contradicting the model output for every row. You can also check out SHAP in python, which is a game theory based model explanation package.
1