What If We Want to Make Predictive Modeling without Target Variable?

[ertr]

Hello everybody,

I try to figure out how can we create a model without a target variable. Do we have to have a Target variable to develop a model? Is it possbile to develop a model without Target? We can use interval Target but are there specific measures for Interval Targets such as it should include three or four distinct values like 1,2,3,4 or nominal like Red, Yellow, Green? Can we elaborate on the definition of Interval Target here?

Also can we create a fake Target?

How can we analyze the data without having a Target variable to interpret the data for future? for forecasting? What are the most commonly used methods? Can Market Basket Analysis be used in this procces?

How can we use Memory Based Reasoning node in this kind of process? Can we also elaborate on the definition of MBA here? We need to understand the logic of MBR well, do you have a samples about this method?

Thanks in advance 🙂

[Reeza]

This is known as an 'unsupervised problem' where you don't know exactly what the output should be. Fraud is a good example, because you often can't be 100% sure if a case is fraud - caught, fraud - not caught/reported, or not fraud. If you search for unsupervised machine learning or problems you'll find many examples and use cases. I don't know of anything specific to SAS though.