Deep learning techniques have recently found applications in the field of predictive business process monitoring. These techniques allow us to predict, among other things, what will be the next events in a case, when will they occur, and which resources will trigger them. They also allow us to generate entire execution traces of a business process, or even entire event logs, which opens up the possibility of using such models for process simulation. This paper addresses the question of how to use deep learning techniques to train accurate models of business process behavior from event logs. The paper proposes an approach to train recurrent neural networks with Long-Short-Term Memory (LSTM) architecture in order to predict sequences of next events, their timestamp, and their associated resource pools. An experimental evaluation on real-life event logs shows that the proposed approach outperforms previously proposed LSTM architectures targeted at this problem.