Remaining time prediction of business pro-cesses plays an important role in resource scheduling and plan making.The structural features of single process instance and the concurrent running of multi-ple process instances are the main factors that affect the accuracy of the remaining time prediction.Ex-isting prediction methods does not take full advan-tage of these two aspects into consideration.To ad-dress this issue,a new prediction method based on trace representation is proposed.More specifically,we first associate the prefix set generated by the event log to different states of the transition system,and en-code the structural features of the prefixes in the state.Then,an annotation containing the feature represen-tation for the prefix and the corresponding remaining time are added to each state to obtain an extended transition system.Next,states in the extended transi-tion system are partitioned by the different lengths of the states,which considers concurrency among mul-tiple process instances.Finally,the long short-term memory (LSTM) deep recurrent neural networks are applied to each partition for predicting the remaining time of new running instances.By extensive experi-mental evaluation using synthetic event logs and real-life event logs,we show that the proposed method out-performs existing baseline methods.