So far we have discussed in a lot of details about building features for Link Prediction models and training these models. Now we need to be able to use such models for making suggestions. Again, think about people you may know banner on Facebook - here we would like to have something similar, like authors you should write a paper with. In addition to that, we will see how to evaluate the model so the evaluation result is more intuitive and clear in this context.
The first step is to re-train the XGBoost model on the entire training set, without the train-validation split. This is easy to do: we just fit the model again on the data before making the split.
Next, we need to process the test dataset. If you remember, the test ...
