I am using the Book recommendation case study to predict the book ratings, the variable rating scale is 1-5, but when the Spark Collaborative filtering Learner node is used to learn the data and Spark Prediction(MLlib) node is used for prediction, the predicted variable value is ranging between -2.5 to 7.7.
Not able to understand why the predicted value is ranging between -2.5 to 7.7 and not between 1 to 5.
Tried normalizing the predicted variable, but still the range remais same.
Sharing the workflow for your reference, i need clarity on the predicted value.
@SK_GL there still is no data attached. In the meantime you could check this example about how such a recommendation system might work:
I think you will have to check how you set up the logic of the recommendation. Maybe you set aside a portion of unknown recommendations by user. Question is if the setup would do this in the current form.
@mlauber71, thanks for sharing the references, but it dint help.
The file size exceeds 4mb, hence unable to upload the workflow. Any other way to upload the work flow or dataset? Tried to upload the datset,but even that file size exceeds 4mb.