topic Re: Help - Issue with Reject Inference Weights in SAS Visual Machine Learning in SAS Data Science

Help - Issue with Reject Inference Weights in SAS Visual Machine Learning

teresacorzo — Tue, 15 Oct 2024 14:29:20 GMT

I’m modelling with reject inference in SAS Visual Machine Learning, I have used the reject inference node with parcelling.

Previously, I’ve created models with reject inference in SAS Miner, where it assigned a weight to every registry in the dataset, assigning 1 to accepted and a given weight between 0 and 1 to the inferred data. However, in SAS Viya, the weights are all 1, in accepted and inferred.

I don’t need all the data in my modelling augmented data, but I don’t want to lose information by creating a random sample that mimics the weight assigned previously in Miner. The weight is given by the following equation:

I’ve tried assing this weight manually, assigning it as a Weight variable, but the node doesn’t read it, it includes all the observations for the rejected data with no calculated weight.

Train set Accepted:

Inferred:

I haven’t found enough information in internet to solve this problem.

Thanks!

Re: Help - Issue with Reject Inference Weights in SAS Visual Machine Learning

Sa_S — Wed, 16 Oct 2024 14:56:13 GMT

Thank you for the post- we are responding in your TS case. Thanks, and have a great day!

Re: Help - Issue with Reject Inference Weights in SAS Visual Machine Learning

RoseCliver — Sat, 16 Nov 2024 06:53:36 GMT

@teresacorzo wrote:
I’m modelling with reject inference in SAS Visual Machine Learning, I have used the reject inference node with parcelling.
Previously, I’ve created models with reject inference in SAS Miner, where it assigned a weight to every registry in the dataset, assigning 1 to accepted and a given weight between 0 and 1 to the inferred data. However, in SAS Viya, the weights are all 1, in accepted and inferred.
I don’t need all the data in my modelling augmented data, but I don’t want to lose information by creating a random sample that mimics the weight assigned previously in Miner. The weight is given by the following equation:

I’ve tried assing this weight manually, assigning it as a Weight variable, but the node doesn’t read it, it includes all the observations for the rejected data with no calculated weight.

Train set Accepted:

Inferred:

I haven’t found enough information in internet to solve this problem.
Thanks!

In SAS Visual Machine Learning, the Reject Inference Node does not automatically assign weights. Manually compute weights as in SAS Miner, then assign them using a custom "Weight" variable in the dataset. Ensure this variable is explicitly defined in the model settings to apply it during training.