Plasmaspheric hiss is a whistler-mode emission that permeates the Earth's plasmasphere and is a significant driver of energetic electron losses through cyclotron resonance. The dynamics of the inner magnetosphere is strongly governed by the interactions between different plasma populations that are coupled through large-scale electric fields. We have developed neural network models that predict Kp from upstream solar wind data. We study the importance of various input parameters, starting with the magnetic component Bz, particle density n, and solar wind speed Vsw. The models are trained on a dataset of 10 years of solar wind and geomagnetic data. We find that the models are able to predict Kp with a high degree of accuracy, and that the inclusion of additional input parameters does not significantly improve the predictions. We also study the relationship between the predicted Kp and the observed geomagnetic activity, and find that the models are able to capture the main features of the geomagnetic response to solar wind variations. Finally, we discuss the implications of our results for the understanding of the inner magnetosphere and the development of predictive models for geomagnetic activity. }