Hi there, well, the predictive power score is the model's ability to achieve some accuracy in general performance, and the feature importance score allows you increase or decrease those accuracy. First of all, you need to select the most important features that allow you achieve a good accuracy in the performance of your model and with that your predictive power score will be great or acceptable 😉

The PPS is an alternative to correlation and thus can be a mean of feature selection among other uses. In that sense, it is not a general score of the model but, just like correlation, it is a score related to the input data that needs data columns as input rather than the model. You can read about the PPS, its application and its familiarity to feature selection here:

https://towardsdatascience.com/rip-correlation-introducing-the-predictive-power-score-3d90808b9598

interested