Infinite insight clustering algorithm

[former_member193862]

Can any one pls help me to understand

-->what is predictive Power and Predictive Confidence in Infinite insight clustering algorithm

--> On what basis do we select the minimum and maximum cluster?

Thanks in Advance!

Regards,

Mirthu

[Former Member · Accepted solution]

Hi Mirthu,

Predictive Power refers to the quality of the data to explain the model. 
This measurement corresponds to the amount of information contained
in the target variable that the explanatory variables are able to explain.

Predictive Confidence refers to the robustness of the model to be applicable
to a new/test data set.  It indicates the capacity of the model to achieve
the same performance when applied to a new data set exhibiting the same characteristics as the training data set.

In clustering models, the observations are grouped depending on their

distance with each centroid.  This can be based on supervised (target
variable) or unsupervised models.  In a supervised cluster model, we
would group like observations that contribute the most predictive power and
confidence, and these observations can be shown to produce the most
influence on the target (such as customers who contribute to revenue).

Hope this helps!

Mike Watschke

SAP Predictive COE

[former_member193862 · Accepted solution]

Thanks a lot Mike.

Basically cluster groups related information based on the quality of data (Power and Confidence). But based on the clusters how do we predict which cluster is the best.

Thanks,

Mirthu

[Former Member · Accepted solution]

Remember that we are just training the model to recognize what makes up a cluster based on centroid approach.  When we actually "apply" the model to a customer or other data set, we are assigning clusterID's based on this learning exercise.  If cluster 7 for example groups customers who bring up the most potential revenue or who bought more than 2 products in a year, we are looking for the same type of customers in a new data set, and assigning them the same cluster 7 designation.   This way, we can target those customers we believe have the highest likelihood to make similar purchases because they meet the training profile.   Hope this helps!

Mike Watschke

SAP Predictive COE