Differential Privacy Preservation for Deep Auto-Encoders: an Application of Human Behavior Prediction